System

Abstract

A system is provided.SOLUTION: A system, comprising: a data storage unit; a preprocessing unit; a model training unit; and a generation unit.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] Current generative AI models learn from a wide range of information available on the internet, making them prone to misinformation and copyright issues. Furthermore, many businesses require reliable data directly linked to specific tasks, rather than general information. However, current AI models do not include proprietary data held by companies, creating problems with fact-checking and data consistency. There is an urgent need to build generative AI models that can resolve these issues and provide highly reliable information directly linked to business. [Means for solving the problem]

[0005] The system of the present invention includes a data storage means, a preprocessing means, a model training means, and a generation means. The data storage means stores academic knowledge data, multilingual data, legal data, general knowledge data, and company-specific data. The preprocessing means reads this data and cleanses and filters it. The model training means trains a text generation model using the preprocessed data. The generation means generates highly accurate responses using the trained model based on prompts from the user. This enables the system of the present invention to quickly provide reliable information directly related to business.

[0006] "Data storage means" refers to a storage device or memory for holding and storing data required by the system.

[0007] "Preprocessing means" refers to a function that performs processing to appropriately cleanse, filter, and format the captured data.

[0008] "Model training means" refers to a function that uses preprocessed data to train machine learning models, particularly text generation models.

[0009] "Generation means" refers to a function that uses a trained text generation model to generate appropriate response text based on input (prompt) from the user.

[0010] "Academic knowledge data" refers to specialized information datasets containing university-level knowledge.

[0011] "Multilingual data" refers to a dataset that contains information in multiple languages.

[0012] "Legal Data" refers to datasets containing information about the laws of different countries.

[0013] "Common knowledge data" refers to datasets that contain general knowledge and information in everyday life and business.

[0014] "Company-specific data" refers to data that is proprietary to a specific company, such as a data set that includes sales data or customer data. [Brief explanation of the drawings]

[0015] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14]FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0016] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0017] First, the terms used in the following description will be explained.

[0018] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0019] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0020] 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.

[0021] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0022] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0023] [First embodiment]

[0024] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0025] 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.

[0026] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0027] 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.

[0028] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0029] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0030] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0031] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0032] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0033] 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.

[0034] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0035] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0036] The system of the present invention includes a data storage means, a pre-processing means, a model training means, and a generation means.

[0037] First, the user configures the system by entering the OpenAI API key, the base data path, and the company's proprietary data path, which then initializes an instance of the CustomAIModule class.

[0038] The server has a data storage means for storing basic data and company-specific data, reads CSV files from each data path, and converts them into data frames. The pre-processing means performs data cleansing and filtering on the read data to generate an integrated dataset.

[0039] The server then runs a model training procedure using the pre-processed data, which includes both baseline data and company-specific data, to create a more accurate text generation model.

[0040] After the model has been trained, a user can input a specific prompt to the system, such as "Please give me some advice on my company's sales strategy." The server tokenizes the prompt and provides it to the trained model. The generator generates an optimal response based on the input prompt and provides the response to the user.

[0041] As a concrete example, consider a usage scenario in a hospital. The user (hospital system administrator) sets the AI's API key, basic data path, and the hospital's proprietary data (e.g., patient treatment data). The server reads the basic data and the hospital's proprietary data, preprocesses them, and creates a training dataset.

[0042] The server trains the GPT-2 model using the training data and displays a message indicating completion of training. When a doctor at the hospital inputs "Please give me advice on the latest treatments" into the generative AI, the server generates information on appropriate treatments from the trained model based on the input prompt and provides it to the doctor.

[0043] In this way, the system of the present invention can quickly provide highly reliable information and realize highly accurate output that is directly linked to specific business operations.

[0044] The processing flow will be explained below.

[0045] Step 1:

[0046] The user sets the OpenAI API key, basic data path, and the company's own data path, and initializes an instance of the CustomAIModule class.

[0047] Step 2:

[0048] The server accesses a data storage means that holds basic data and company-specific data, and reads CSV files from each data path.

[0049] Step 3:

[0050] The server uses preprocessing tools to cleanse the ingested data by handling missing values ​​and removing unnecessary columns.

[0051] Step 4:

[0052] The server combines the base data and the company's proprietary data into one integrated data set, so that all data is in a unified format.

[0053] Step 5:

[0054] The server executes a model training means to train a text generation model using the preprocessed data, which involves extracting and tokenizing text data from the dataset.

[0055] Step 6:

[0056] The server begins training the text generation model using the tokenized text data. The training process is set up over multiple epochs, during which the model learns the features of the text.

[0057] Step 7:

[0058] The server confirms that training is complete and displays a message to the user saying "Model training is complete."

[0059] Step 8:

[0060] The user inputs a specific prompt (e.g., "Please give me some advice on our company's sales strategy") to the system.

[0061] Step 9:

[0062] The server tokenizes the input prompt and feeds it into a trained text generation model.

[0063] Step 10:

[0064] The server uses the trained model to generate text responses based on the prompts, ensuring accurate and relevant responses.

[0065] Step 11:

[0066] The server decodes the generated response text and presents it to the user.

[0067] Step 12:

[0068] The user checks the response text provided and uses it in their work as needed.

[0069] Example 1

[0070] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0071] In information generation systems using conventional generative AI models, it has been difficult to train highly accurate models that combine reliable data with diverse data sources, and it has also been challenging to quickly and appropriately provide optimal responses to prompts from users. In particular, when customized information is required for each company or industry, a method for efficiently handling training data that matches that needs has been required.

[0072] 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.

[0073] In this invention, the server includes a data storage unit, a preprocessing unit, a model training unit, a generation unit, a user setting input unit, a prompt input unit, a response generation unit, and a presentation unit. This allows for the effective integration of a diverse set of reliable data to train a highly accurate generative AI model customized for each company or industry. This allows for the provision of prompt and appropriate responses to specific prompts from users.

[0074] "Data storage means" refers to a means for storing various data including basic data and company-specific data provided by users.

[0075] The "pre-processing means" is a means for performing pre-processing such as cleansing, filtering, and integration on the data read from the data storage means.

[0076] "Model training means" means a means for training a generative AI model using preprocessed data.

[0077] "Generation means" means a means for generating an optimal response to a prompt input by a user using a trained generative AI model.

[0078] The "user setting input means" is a means by which a user inputs an API key, a basic data path, and a company-specific data path into the system.

[0079] A "prompt input means" is a means by which a user inputs specific questions or requests into the system.

[0080] A "response generation means" is a means for tokenizing an input prompt and generating a response using a trained generative AI model.

[0081] A "presentation means" is a means for displaying the generated response to the user.

[0082] The system of the present invention includes a data storage means, a preprocessing means, a model training means, a generation means, a user setting input means, a prompt input means, a response generation means, and a presentation means.

[0083] First, the user configures the system by entering an API key, a base data path, and a company-specific data path. This configuration initializes an instance of the CustomAIModule class. The data path entered by the user contains customizable data for each company.

[0084] Next, the server stores the basic data and company-specific data through a data storage means. The server reads the CSV file from the data path specified by the user and converts it into a data frame using the Pandas library.

[0085] A preprocessing tool then performs data cleansing and filtering on the data frames, including handling missing values, removing unnecessary columns, and filtering the data to produce a single unified dataset.

[0086] The server then runs a model training procedure using the combined dataset. The training dataset includes both baseline and company-specific data, resulting in a more accurate generative AI model. This training utilizes deep learning frameworks such as TensorFlow and PyTorch. Training can take anywhere from several hours to several days.

[0087] Once the model is trained, a user can enter a specific prompt through the prompt input means, for example, "Please give me some advice on my company's sales strategy." The server tokenizes this prompt and feeds the preprocessed tokens to the generative AI model.

[0088] The response generation means uses the trained model to generate an optimal response based on the input prompt. This response is provided to the user through the presentation means. For example, in response to the prompt "Please give me some advice on my company's sales strategy," the trained model generates advice on an appropriate sales strategy and provides it to the user.

[0089] As a concrete example, consider a usage scenario in a hospital. The user (hospital system administrator) sets the API key, basic data path, and hospital-specific data path (e.g., patient treatment data). The server reads the basic data and the hospital-specific data, preprocesses them, and creates an integrated dataset. The server trains the generative AI model using the training data and displays a message indicating training completion. When a doctor at the hospital inputs to the generative AI, "Please give me some advice on the latest treatments," the server generates information on the optimal treatment based on the input prompt and provides it to the doctor.

[0090] In this way, the system of the present invention can quickly provide highly reliable information and realize highly accurate output that is directly linked to specific business operations.

[0091] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0092] Step 1:

[0093] User-configured system

[0094] Input: API key, basic data path, company's own data path

[0095] Description: The user enters the API key, basic data path, and company-specific data path in the system settings screen. This setting registers the necessary information in the system and initializes an instance of the CustomAIModule class.

[0096] Output: An initialized instance of the CustomAIModule class.

[0097] Step 2:

[0098] Data loading

[0099] Input: User specified data path

[0100] Description: The server uses the data storage means to read the CSV file from the basic data path and company-specific data path entered by the user. At this time, the CSV file is converted to a data frame using the Pandas library and stored in memory.

[0101] Output: Basic and company-specific data in a data frame format

[0102] Step 3:

[0103] Data Preprocessing

[0104] Input: Basic data and company-specific data in a data frame format

[0105] Description: The preprocessing method performs data cleansing and filtering on the data frame, specifically handling missing values, removing unnecessary columns, and filtering the data to produce a single unified dataset.

[0106] Output: A consolidated dataset

[0107] Step 4:

[0108] Model Training

[0109] Input: Integrated dataset

[0110] Description: The server runs a model training procedure using the combined dataset. The training dataset includes both baseline and company-specific data, resulting in a more accurate generative AI model. This training utilizes deep learning frameworks such as TensorFlow and PyTorch. Training can take anywhere from several hours to several days.

[0111] Output: A trained generative AI model

[0112] Step 5:

[0113] Prompt Input

[0114] Input: A specific prompt from the user (e.g., "Please give me some advice on our company's sales strategy.")

[0115] Description: The user enters a specific prompt into the system, which contains information needed to generate a future response.

[0116] Output: The prompt entered

[0117] Step 6:

[0118] Prompt Tokenization

[0119] Input: Entered prompt

[0120] Description: The server tokenizes the input prompt by splitting it into words or sub-words and converting each into a token, often using Hugging Face's Transformers library.

[0121] Output: Tokenized prompt

[0122] Step 7:

[0123] Response Generation

[0124] Input: Tokenized prompt, trained generative AI model

[0125] Description: The response generator uses a trained model to generate optimal responses to tokenized prompts.

[0126] Output: The generated response

[0127] Step 8:

[0128] Response suggestion

[0129] Input: Generated response

[0130] Description: The generated response is provided to the user through a presentation method. The response is displayed on the user's device and used as necessary information.

[0131] Output: The response displayed on the user's terminal

[0132] (Application example 1)

[0133] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0134] In recent years, customer targeting has become more sophisticated in the advertising industry, creating a demand for efficient ad generation that utilizes a variety of data. However, traditional ad generation methods require manual copywriting and targeting, which is time-consuming and labor-intensive, and it is difficult to generate consistent ad copy. To solve this problem, a system is needed that utilizes corporate and market data and automatically generates accurate ad copy based on user prompts.

[0135] 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.

[0136] In this invention, the server includes a data storage means, a preprocessing means, a model training means, and an advertising copy generation means, which preprocesses company data and market data and enables the automatic generation of consistent and highly accurate advertising copy.

[0137] "Data storage means" means means for holding basic data, company-specific data, academic knowledge data, multilingual data, legal data, general knowledge data and market data.

[0138] The "preprocessing means" is a means for reading data from the data storage means, processing missing values, deleting unnecessary columns, and filtering the data.

[0139] A "model training means" is a means for training a natural language processing model using preprocessed data and constructing a generative AI model.

[0140] "Generating means" means using a trained generative AI model to generate an optimal response based on prompts from a user.

[0141] The "advertising copy generation means" is a means for automatically generating consistent and highly accurate advertising copy based on company data and market data in response to user prompts.

[0142] "Company Data" means proprietary data related to a specific company that is used for targeting and personalization in advertising generation.

[0143] "Market Data" means data related to a particular market or industry and used for trend analysis and targeting.

[0144] This invention is a system for automatically generating advertising copy, comprising a data storage means, a preprocessing means, a model training means, a generation means, and an advertising copy generation means, which utilizes company data and market data to generate accurate advertising copy based on specific prompts from a user.

[0145] Data Storage Means

[0146] The server includes data storage means for holding basic data, company-specific data, academic knowledge data, multilingual data, legal data, general knowledge data, and market data. This data storage means includes hard disk drives and database systems required for data capture and management.

[0147] Pretreatment means

[0148] The server has a preprocessing means for reading data from the data storage means, processing missing values, deleting unnecessary items, filtering data, etc. For example, the server uses a data processing tool such as the Pandas library to efficiently preprocess the data.

[0149] Model training methods

[0150] Using the preprocessed data, a model training method builds a generative AI model, which uses a machine learning framework such as Python's Transformers library or TensorFlow for training.

[0151] generation means

[0152] The generator uses a trained generative AI model to generate an optimal response based on the user's prompts. Specifically, it uses the OpenAI API to invoke a natural language processing model and generate text based on the user's request.

[0153] Ad copy generation method

[0154] The advertising copy generation means automatically generates targeted advertising copy based on a prompt text provided by a user, utilizing company data and market data.

[0155] For example, if a user inputs a prompt such as "Please write an ad copy for a new dress from a fashion brand aimed at women in their 20s" on their smartphone or computer, the server tokenizes the prompt and feeds it to the trained model. The generated ad copy is then displayed.

[0156] For example, consider the following prompt:

[0157] "Please write an ad copy for a new dress from a fashion brand aimed at women in their 20s."

[0158] This allows users to quickly generate consistent and accurate ad copy and effectively develop advertising campaigns. In this way, the system of the present invention significantly improves the efficiency and accuracy of ad generation.

[0159] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0160] Step 1:

[0161] A user accesses the system and enters the API key and data path.

[0162] Input: API key, basic data path, corporate data path

[0163] Output: An initialized CustomAIModule instance.

[0164] How it works: A user logs into the system from a smartphone or PC and enters their OpenAI API key and data path. The server receives this information and initializes the instance.

[0165] Step 2:

[0166] The server reads the data from the data storage means.

[0167] Input: Basic Data Path, Corporate Data Path

[0168] Output: Data frame

[0169] What it does: The server uses the Pandas library to read the CSV file from the specified path and convert the base and company data into a dataframe.

[0170] Step 3:

[0171] The server executes the preprocessing means to process the data.

[0172] Input: DataFrame

[0173] Output: Preprocessed dataset

[0174] What it does: The server preprocesses the data frame, handling missing values, removing unnecessary items, and filtering the data, resulting in a clean, unified dataset.

[0175] Step 4:

[0176] The server uses the preprocessed data to perform model training.

[0177] Input: Preprocessed dataset

[0178] Output: A trained generative AI model

[0179] How it works: The server uses the Transformers library and TensorFlow to train a generative AI model using a preprocessed dataset. After training is complete, a highly accurate generative AI model is created.

[0180] Step 5:

[0181] The user enters a prompt statement.

[0182] Input: prompt statement

[0183] Output: Tokenized prompt

[0184] How it works: The user inputs a specific prompt, such as "Please write an ad copy for a new dress from a fashion brand aimed at women in their 20s," from their smartphone or computer. The server receives this prompt and performs a tokenization process.

[0185] Step 6:

[0186] The server generates the advertisement copy using the generation means.

[0187] Input: Tokenized prompt

[0188] Output: Generated ad text

[0189] How it works: The server feeds the tokenized prompts to a trained generative AI model to generate optimal ad copy, which is then displayed to the user.

[0190] Step 7:

[0191] The user reviews and uses the generated ad copy.

[0192] Input: Generated ad text

[0193] Output: Available ad text

[0194] How it works: Users review the ad copy provided by the system, modify or edit it as needed, and use it as the final ad copy, allowing for efficient and consistent ad campaigns.

[0195] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0196] The system of the present invention includes a data storage means, a pre-processing means, a model training means, a generation means, and an emotion engine.

[0197] First, the user configures the system by entering the OpenAI API key, the base data path, and the company's proprietary data path, and then initializes an instance of the CustomAIModule class. This gets the system ready to run.

[0198] The server accesses the data storage means that holds the basic data and the company's proprietary data, and reads the CSV files from each data path. The pre-processing means performs data cleansing and filtering on the data to generate an integrated data set.

[0199] The server then executes a model training process using the preprocessed data, which includes the basic data, the company-specific data, and the emotion information data, to create a highly accurate text generation model with emotion recognition capabilities.

[0200] After the model is trained, users can input specific prompts into the system, such as "Please give me some advice on how to respond to a customer complaint." The server tokenizes the prompt and provides it to the emotion engine, which analyzes the prompt content and classifies the user's emotion.

[0201] The server then feeds the emotion information classified by the emotion engine to a trained text generation model, and uses a generation means to generate an optimal response text, which includes an appropriate response according to the emotion.

[0202] As a concrete example, consider a customer support scenario. The user (customer support representative) sets the AI's API key, basic data path, and company-specific data (e.g., past customer interaction history). The server reads and preprocesses the basic data, company-specific data, and past customer interaction data to create a training dataset.

[0203] The server trains the model using the training data and displays a message indicating completion of training. When a customer support representative inputs "What should I do if a customer is angry about a product?" into the generative AI, the server analyzes the input prompt with an emotion engine and classifies the emotion. The server generates a response text according to the emotion and provides it to the representative.

[0204] In this way, the system of the present invention is able to recognize the user's emotions and quickly provide highly accurate and reliable information in response to the emotions.

[0205] The processing flow will be explained below.

[0206] Step 1:

[0207] The user sets the OpenAI API key, basic data path, and the company's own data path, and initializes an instance of the CustomAIModule class.

[0208] Step 2:

[0209] The server accesses a data storage means that holds basic data and company-specific data, and reads CSV files from each data path.

[0210] Step 3:

[0211] The server uses preprocessing tools to perform data cleansing on the loaded data, including handling missing values ​​and removing unnecessary columns.

[0212] Step 4:

[0213] The server combines the base data and the company's proprietary data into one integrated data set, so that all data is in a unified format.

[0214] Step 5:

[0215] The server executes a model training means to train a text generation model using the preprocessed data, which involves extracting and tokenizing text data from the dataset.

[0216] Step 6:

[0217] The server adds emotion information to the training data and creates a new training set that includes information corresponding to emotions.

[0218] Step 7:

[0219] The server begins training the text generation model using the tokenized text data and sentiment information. The training process is configured to span multiple epochs, during which the model learns the association between text features and sentiment.

[0220] Step 8:

[0221] The server confirms that training is complete and displays a message to the user saying "Model training is complete."

[0222] Step 9:

[0223] The user enters a specific prompt into the system (e.g., "What should I do if a customer is upset about my product?").

[0224] Step 10:

[0225] The server tokenizes the input prompt and feeds it to the emotion engine, which analyzes the user's input and classifies the emotion.

[0226] Step 11:

[0227] The server supplies the emotion information classified by the emotion engine to a trained text generation model, and generates an optimal response text using a generation means.

[0228] Step 12:

[0229] The server decodes the generated response text and presents it to the user.

[0230] Step 13:

[0231] The user checks the response text provided and uses it in their work as needed.

[0232] Example 2

[0233] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0234] Currently, there is a lack of systems that can accurately recognize customer emotions and quickly provide appropriate responses. This results in a problem of reduced efficiency of human resources and time in customer service operations in various fields, where efficient and effective support is required. In light of this situation, there is a need to develop a system that can achieve highly accurate emotion recognition and response generation.

[0235] 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.

[0236] In this invention, the server includes a user setting means, a data storage means, a preprocessing means, a model training means, a sentiment analysis means, and a text generation means, which enable the server to recognize the user's sentiment and quickly provide highly accurate and reliable information in response to the sentiment.

[0237] The "user setting means" is a function that allows the user to input setting information required to start the system operation.

[0238] "Data storage means" refers to a storage device or system for holding basic data, company-specific data, and sentiment information data.

[0239] The "preprocessing means" is a function that performs missing value complementation, correction of inconsistent data, and filtering of the acquired data to generate a clean dataset.

[0240] The "model training means" is a function that uses preprocessed data to train a machine learning model and create a highly accurate text generation model.

[0241] The "emotion analysis means" is a function that analyzes the user's emotions from the input text data and assigns appropriate emotion labels.

[0242] The "text generation means" is a function that generates optimal response text based on the emotion data obtained by the emotion analysis means.

[0243] The system of the present invention realizes highly accurate text generation by allowing the user to configure the system and input specific prompt sentences. Specifically, the system includes the following means: user configuration means, data storage means, preprocessing means, model training means, sentiment analysis means, and text generation means.

[0244] First, the user enters the necessary configuration information to set up the system, including the API key, the base data path, and the company's own data path. This initializes an instance of the CustomAIModule class, and the system is ready to run.

[0245] The server accesses the data storage means and reads the basic data, company-specific data, and emotion information data in CSV format. The data storage means refers to a storage device for storing these data.

[0246] Next, the server preprocesses the data using a preprocessing method, which performs missing value imputation, inconsistent data correction, duplicate data removal, and data filtering to generate a clean dataset. Specifically, the server performs data cleansing and filtering on the loaded data.

[0247] The server then uses a model training means to train a machine learning model based on the preprocessed data. The training data includes basic data, company-specific data, and sentiment information data. The training process is repeated to build a highly accurate text generation model. When training is complete, the server displays a message indicating that training is complete.

[0248] The user can input a specific prompt to the system. For example, "Please give me some advice on how to respond to a customer complaint." The server tokenizes this input and provides it to the sentiment analyzer. The sentiment analyzer analyzes the prompt content and classifies the user's sentiment.

[0249] Next, the server supplies the emotion information obtained by the emotion analysis means to the trained text generation model, and uses the text generation means to generate an optimal response text, which includes an appropriate response according to the emotion.

[0250] A concrete example of this would be a customer support scenario. The user (customer support representative) sets an API key, a basic data path, and the company's proprietary data (e.g., past customer interaction history). The server loads and preprocesses the basic data, the company's proprietary data, and past customer interaction data to create a training dataset. The server then trains the model using the training data, and a message indicating training completion is displayed. When the customer support representative inputs the following into the generative AI: "What should I do if a customer is upset about a product?", the server analyzes this input using the emotion engine and assigns an appropriate emotion label. Finally, the server generates the optimal response text based on the emotion and provides it to the representative.

[0251] In this way, the system of the present invention is able to recognize the user's emotions and quickly provide highly accurate and reliable information in response to the emotions.

[0252] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0253] Step 1:

[0254] The user configures the system.

[0255] Input: OpenAI API key, basic data path, company's proprietary data path

[0256] Operation: The user enters this information on the screen and sends it to the system. An instance of the CustomAIModule class is initialized and the system is ready to operate.

[0257] Output: System completed initialization

[0258] Step 2:

[0259] The server accesses the data storage means and reads the data.

[0260] Input: Basic data path, company's proprietary data path

[0261] Operation: Accesses the data storage means and reads the CSV files from each data path.

[0262] Output: Raw dataset of loaded basic and company data

[0263] Step 3:

[0264] The server pre-processes the data using the pre-processing means.

[0265] Input: Raw dataset

[0266] What it does: It performs data cleansing and filtering on the data it loads, including imputing missing values, correcting inconsistent data, removing duplicates, and filtering data.

[0267] Output: A clean dataset

[0268] Step 4:

[0269] The server trains the model using the model training means.

[0270] Input: Clean dataset, basic data, company data, sentiment data

[0271] How it works: A machine learning algorithm is used to train a model. The training process is repeated to build a highly accurate text generation model.

[0272] Output: A fully trained, highly accurate text generation model

[0273] Step 5:

[0274] The user enters a specific prompt.

[0275] Input: Prompt text (e.g., "Please advise how to handle a customer complaint.")

[0276] Action: Sends the entered prompt to the system.

[0277] Output: The prompt text entered into the system

[0278] Step 6:

[0279] The server parses the prompt.

[0280] Input: prompt statement

[0281] How it works: The prompt sentence is tokenized and fed to the sentiment analyzer, which analyzes the prompt content and classifies the user's sentiment.

[0282] Output: Classified emotion data

[0283] Step 7:

[0284] The server generates a response text based on the emotion information.

[0285] Input: Classified emotion data, trained text generation model

[0286] Operation: Generate optimal response text based on emotional information obtained by the emotion analysis means. Generate highly accurate text using the generation means.

[0287] Output: Optimal response text depending on the sentiment

[0288] Step 8:

[0289] The server provides the response text to the user.

[0290] Input: Response text

[0291] Behavior: The server generates a response text and displays it on the user's screen.

[0292] Output: The response text provided to the user

[0293] (Application example 2)

[0294] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0295] Conventional customer support systems have difficulty in recognizing and responding to users' emotions appropriately, making it difficult to increase user satisfaction, especially on online shopping sites. Returning a uniform response without considering emotions can further increase user dissatisfaction.

[0296] 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.

[0297] In this invention, the server includes a data storage means, a preprocessing means, a model training means, a user response generation means, and an emotion recognition means, which enable the server to analyze prompts input by a user, appropriately recognize the emotion, and generate a highly accurate response accordingly.

[0298] The "data storage means" is a storage device for storing data entered by users, academic knowledge data, multilingual data, legal data, general knowledge data, and company-specific data.

[0299] The "preprocessing means" is a device or software that performs cleansing such as processing missing values, deleting unnecessary columns, and filtering data on the data acquired from the data storage means, and integrates the data.

[0300] A "model training means" is a device or software for training a text generation model using preprocessed data.

[0301] A "generator" is a device or software that uses a trained model to generate response text based on a user's input.

[0302] An "emotion recognition means" is a device or software that analyzes the prompt entered by the user and classifies the emotion from its content.

[0303] The "user response generation means" is a device or software that generates an optimal response text based on the emotion information classified by the emotion recognition means.

[0304] A "prompt" is a question or request that a user enters into a system.

[0305] A "training dataset" is a collection of data that has been preprocessed for training a model.

[0306] "Emotion information" is information that indicates the type and intensity of the emotion contained in the user's prompt, as analyzed by the emotion recognition means.

[0307] A "text generation model" is a model that learns from a training dataset to generate appropriate text responses based on user prompts.

[0308] The system of the present invention includes a data storage means, a preprocessing means, a model training means, a generation means, an emotion recognition means, and a user response generation means. This system is specifically implemented as follows.

[0309] The system of the present invention first stores basic data entered by users, company-specific data, and emotion information data in a data storage means. These data are often saved as CSV files.

[0310] Next, the preprocessing means reads the data from the data storage means, handles missing values, removes unnecessary columns, and filters the data to generate a unified dataset, which provides consistent data for subsequent training processes.

[0311] The model training method uses preprocessed data to train a generative AI model. This training includes basic data, company-specific data, and sentiment data. The program is typically implemented using an existing language model API, such as the OpenAI API.

[0312] Once training is complete, the user response generator operates and accepts prompts from the user. For example, if the user inputs "A customer is angry about the product. What should I do?", the system analyzes the prompt with the emotion recognition unit and classifies the user's emotion.

[0313] The emotion information recognized by the emotion recognition means is supplied to the generation means, which generates an appropriate response text. The generated response is in line with the user's emotion and is provided with high accuracy and speed.

[0314] Specifically, when a user uses a customer support application on their smartphone and inputs, "My order hasn't arrived yet. What's going on?", the system first analyzes the input using emotion recognition and recognizes the emotion of "dissatisfaction (anger)." It then uses a generative AI model to generate an appropriate response text, providing the user with a support message such as, "We apologize for the delay in your order. We will check the current situation and get back to you as soon as possible."

[0315] Therefore, the system of the present invention can solve the problems that conventional systems have had and provide highly accurate customer support that responds to the user's emotions.

[0316] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0317] Step 1:

[0318] A user inputs a prompt sentence through a smartphone application, for example, "A customer is angry about the product. What should I do?" This input is sent to the server in text format.

[0319] Step 2:

[0320] The server reads the necessary data from the data storage means, which includes basic data, company-specific data, and sentiment data. The data is stored as CSV files, and the server parses these files.

[0321] Input: CSV file stored in your data storage means

[0322] Output: Raw data (basic data, company-specific data, sentiment data)

[0323] Step 3:

[0324] Preprocessing procedures are used on the raw data to handle missing values, remove unnecessary columns, and filter the data, creating a consistent, integrated dataset.

[0325] Input: Raw data read

[0326] Output: A consolidated dataset

[0327] Step 4:

[0328] The server uses a model training method to train a generative AI model using the pre-processed dataset, which is often done using an external language model API such as the OpenAI API.

[0329] Input: Preprocessed dataset

[0330] Output: A trained generative AI model

[0331] Step 5:

[0332] The server receives the prompt sentence entered by the user and analyzes it with the emotion recognition means, which analyzes the text of the prompt sentence and classifies the user's emotion.

[0333] Input: User prompt text

[0334] Output: Classified emotion information

[0335] Step 6:

[0336] The classified emotion information is supplied to the server's generation means, which generates a response text that matches the emotion. The generation means uses a trained generative AI model to generate the optimal response sentence.

[0337] Input: Classified emotion information, user prompt

[0338] Output: The generated response text

[0339] Step 7:

[0340] The response text generated by the response generating means is sent to the user terminal, and the user can view the generated response through a smartphone application.

[0341] Input: Generated response text

[0342] Output: The response message that is displayed to the user

[0343] The above steps enable highly accurate customer support that takes into account the user's emotions.

[0344] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0345] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0346] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0347] [Second embodiment]

[0348] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0349] 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.

[0350] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0351] 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.

[0352] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0353] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0354] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0355] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0356] 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.

[0357] 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.

[0358] 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.

[0359] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0360] The system of the present invention includes a data storage means, a pre-processing means, a model training means, and a generation means.

[0361] First, the user configures the system by entering the OpenAI API key, the base data path, and the company's proprietary data path, which then initializes an instance of the CustomAIModule class.

[0362] The server has a data storage means for storing basic data and company-specific data, reads CSV files from each data path, and converts them into data frames. The pre-processing means performs data cleansing and filtering on the read data to generate an integrated dataset.

[0363] The server then runs a model training procedure using the pre-processed data, which includes both baseline data and company-specific data, to create a more accurate text generation model.

[0364] After the model has been trained, a user can input a specific prompt to the system, such as "Please give me some advice on my company's sales strategy." The server tokenizes the prompt and provides it to the trained model. The generator generates an optimal response based on the input prompt and provides the response to the user.

[0365] As a concrete example, consider a usage scenario in a hospital. The user (hospital system administrator) sets the AI's API key, basic data path, and the hospital's proprietary data (e.g., patient treatment data). The server reads the basic data and the hospital's proprietary data, preprocesses them, and creates a training dataset.

[0366] The server trains the GPT-2 model using the training data and displays a message indicating completion of training. When a doctor at the hospital inputs "Please give me advice on the latest treatments" into the generative AI, the server generates information on appropriate treatments from the trained model based on the input prompt and provides it to the doctor.

[0367] In this way, the system of the present invention can quickly provide highly reliable information and realize highly accurate output that is directly linked to specific business operations.

[0368] The processing flow will be explained below.

[0369] Step 1:

[0370] The user sets the OpenAI API key, basic data path, and the company's own data path, and initializes an instance of the CustomAIModule class.

[0371] Step 2:

[0372] The server accesses a data storage means that holds basic data and company-specific data, and reads CSV files from each data path.

[0373] Step 3:

[0374] The server uses preprocessing tools to cleanse the ingested data by handling missing values ​​and removing unnecessary columns.

[0375] Step 4:

[0376] The server combines the base data and the company's proprietary data into one integrated data set, so that all data is in a unified format.

[0377] Step 5:

[0378] The server executes a model training means to train a text generation model using the preprocessed data, which involves extracting and tokenizing text data from the dataset.

[0379] Step 6:

[0380] The server begins training the text generation model using the tokenized text data. The training process is set up over multiple epochs, during which the model learns the features of the text.

[0381] Step 7:

[0382] The server confirms that training is complete and displays a message to the user saying "Model training is complete."

[0383] Step 8:

[0384] The user inputs a specific prompt (e.g., "Please give me some advice on our company's sales strategy") to the system.

[0385] Step 9:

[0386] The server tokenizes the input prompt and feeds it into a trained text generation model.

[0387] Step 10:

[0388] The server uses the trained model to generate text responses based on the prompts, ensuring accurate and relevant responses.

[0389] Step 11:

[0390] The server decodes the generated response text and presents it to the user.

[0391] Step 12:

[0392] The user checks the response text provided and uses it in their work as needed.

[0393] Example 1

[0394] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0395] In information generation systems using conventional generative AI models, it has been difficult to train highly accurate models that combine reliable data with diverse data sources, and it has also been challenging to quickly and appropriately provide optimal responses to prompts from users. In particular, when customized information is required for each company or industry, a method for efficiently handling training data that matches that needs has been required.

[0396] 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.

[0397] In this invention, the server includes a data storage unit, a preprocessing unit, a model training unit, a generation unit, a user setting input unit, a prompt input unit, a response generation unit, and a presentation unit. This allows for the effective integration of a diverse set of reliable data to train a highly accurate generative AI model customized for each company or industry. This allows for the provision of prompt and appropriate responses to specific prompts from users.

[0398] "Data storage means" refers to a means for storing various data including basic data and company-specific data provided by users.

[0399] The "pre-processing means" is a means for performing pre-processing such as cleansing, filtering, and integration on the data read from the data storage means.

[0400] "Model training means" means a means for training a generative AI model using preprocessed data.

[0401] "Generation means" means a means for generating an optimal response to a prompt input by a user using a trained generative AI model.

[0402] The "user setting input means" is a means by which a user inputs an API key, a basic data path, and a company-specific data path into the system.

[0403] A "prompt input means" is a means by which a user inputs specific questions or requests into the system.

[0404] A "response generation means" is a means for tokenizing an input prompt and generating a response using a trained generative AI model.

[0405] A "presentation means" is a means for displaying the generated response to the user.

[0406] The system of the present invention includes a data storage means, a preprocessing means, a model training means, a generation means, a user setting input means, a prompt input means, a response generation means, and a presentation means.

[0407] First, the user configures the system by entering an API key, a base data path, and a company-specific data path. This configuration initializes an instance of the CustomAIModule class. The data path entered by the user contains customizable data for each company.

[0408] Next, the server stores the basic data and company-specific data through a data storage means. The server reads the CSV file from the data path specified by the user and converts it into a data frame using the Pandas library.

[0409] A preprocessing tool then performs data cleansing and filtering on the data frames, including handling missing values, removing unnecessary columns, and filtering the data to produce a single unified dataset.

[0410] The server then runs a model training procedure using the combined dataset. The training dataset includes both baseline and company-specific data, resulting in a more accurate generative AI model. This training utilizes deep learning frameworks such as TensorFlow and PyTorch. Training can take anywhere from several hours to several days.

[0411] Once the model is trained, a user can enter a specific prompt through the prompt input means, for example, "Please give me some advice on my company's sales strategy." The server tokenizes this prompt and feeds the preprocessed tokens to the generative AI model.

[0412] The response generation means uses the trained model to generate an optimal response based on the input prompt. This response is provided to the user through the presentation means. For example, in response to the prompt "Please give me some advice on my company's sales strategy," the trained model generates advice on an appropriate sales strategy and provides it to the user.

[0413] As a concrete example, consider a usage scenario in a hospital. The user (hospital system administrator) sets the API key, basic data path, and hospital-specific data path (e.g., patient treatment data). The server reads the basic data and the hospital-specific data, preprocesses them, and creates an integrated dataset. The server trains the generative AI model using the training data and displays a message indicating training completion. When a doctor at the hospital inputs to the generative AI, "Please give me some advice on the latest treatments," the server generates information on the optimal treatment based on the input prompt and provides it to the doctor.

[0414] In this way, the system of the present invention can quickly provide highly reliable information and realize highly accurate output that is directly linked to specific business operations.

[0415] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0416] Step 1:

[0417] User-configured system

[0418] Input: API key, basic data path, company's own data path

[0419] Description: The user enters the API key, basic data path, and company-specific data path in the system settings screen. This setting registers the necessary information in the system and initializes an instance of the CustomAIModule class.

[0420] Output: An initialized instance of the CustomAIModule class.

[0421] Step 2:

[0422] Data loading

[0423] Input: User specified data path

[0424] Description: The server uses the data storage means to read the CSV file from the basic data path and company-specific data path entered by the user. At this time, the CSV file is converted to a data frame using the Pandas library and stored in memory.

[0425] Output: Basic and company-specific data in a data frame format

[0426] Step 3:

[0427] Data Preprocessing

[0428] Input: Basic data and company-specific data in a data frame format

[0429] Description: The preprocessing method performs data cleansing and filtering on the data frame, specifically handling missing values, removing unnecessary columns, and filtering the data to produce a single unified dataset.

[0430] Output: A consolidated dataset

[0431] Step 4:

[0432] Model Training

[0433] Input: Integrated dataset

[0434] Description: The server runs a model training procedure using the combined dataset. The training dataset includes both baseline and company-specific data, resulting in a more accurate generative AI model. This training utilizes deep learning frameworks such as TensorFlow and PyTorch. Training can take anywhere from several hours to several days.

[0435] Output: A trained generative AI model

[0436] Step 5:

[0437] Prompt Input

[0438] Input: A specific prompt from the user (e.g., "Please give me some advice on our company's sales strategy.")

[0439] Description: The user enters a specific prompt into the system, which contains information needed to generate a future response.

[0440] Output: The prompt entered

[0441] Step 6:

[0442] Prompt Tokenization

[0443] Input: Entered prompt

[0444] Description: The server tokenizes the input prompt by splitting it into words or sub-words and converting each into a token, often using Hugging Face's Transformers library.

[0445] Output: Tokenized prompt

[0446] Step 7:

[0447] Response Generation

[0448] Input: Tokenized prompt, trained generative AI model

[0449] Description: The response generator uses a trained model to generate optimal responses to tokenized prompts.

[0450] Output: The generated response

[0451] Step 8:

[0452] Response suggestion

[0453] Input: Generated response

[0454] Description: The generated response is provided to the user through a presentation method. The response is displayed on the user's device and used as necessary information.

[0455] Output: The response displayed on the user's terminal

[0456] (Application example 1)

[0457] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0458] In recent years, customer targeting has become more sophisticated in the advertising industry, creating a demand for efficient ad generation that utilizes a variety of data. However, traditional ad generation methods require manual copywriting and targeting, which is time-consuming and labor-intensive, and it is difficult to generate consistent ad copy. To solve this problem, a system is needed that utilizes corporate and market data and automatically generates accurate ad copy based on user prompts.

[0459] 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.

[0460] In this invention, the server includes a data storage means, a preprocessing means, a model training means, and an advertising copy generation means, which preprocesses company data and market data and enables the automatic generation of consistent and highly accurate advertising copy.

[0461] "Data storage means" means means for holding basic data, company-specific data, academic knowledge data, multilingual data, legal data, general knowledge data and market data.

[0462] The "preprocessing means" is a means for reading data from the data storage means, processing missing values, deleting unnecessary columns, and filtering the data.

[0463] A "model training means" is a means for training a natural language processing model using preprocessed data and constructing a generative AI model.

[0464] "Generating means" means using a trained generative AI model to generate an optimal response based on prompts from a user.

[0465] The "advertising copy generation means" is a means for automatically generating consistent and highly accurate advertising copy based on company data and market data in response to user prompts.

[0466] "Company Data" means proprietary data related to a specific company that is used for targeting and personalization in advertising generation.

[0467] "Market Data" means data related to a particular market or industry and used for trend analysis and targeting.

[0468] This invention is a system for automatically generating advertising copy, comprising a data storage means, a preprocessing means, a model training means, a generation means, and an advertising copy generation means, which utilizes company data and market data to generate accurate advertising copy based on specific prompts from a user.

[0469] Data Storage Means

[0470] The server includes data storage means for holding basic data, company-specific data, academic knowledge data, multilingual data, legal data, general knowledge data, and market data. This data storage means includes hard disk drives and database systems required for data capture and management.

[0471] Pretreatment means

[0472] The server has a preprocessing means for reading data from the data storage means, processing missing values, deleting unnecessary items, filtering data, etc. For example, the server uses a data processing tool such as the Pandas library to efficiently preprocess the data.

[0473] Model training methods

[0474] Using the preprocessed data, a model training method builds a generative AI model, which uses a machine learning framework such as Python's Transformers library or TensorFlow for training.

[0475] generation means

[0476] The generator uses a trained generative AI model to generate an optimal response based on the user's prompts. Specifically, it uses the OpenAI API to invoke a natural language processing model and generate text based on the user's request.

[0477] Ad copy generation method

[0478] The advertising copy generation means automatically generates targeted advertising copy based on a prompt text provided by a user, utilizing company data and market data.

[0479] For example, if a user inputs a prompt such as "Please write an ad copy for a new dress from a fashion brand aimed at women in their 20s" on their smartphone or computer, the server tokenizes the prompt and feeds it to the trained model. The generated ad copy is then displayed.

[0480] For example, consider the following prompt:

[0481] "Please write an ad copy for a new dress from a fashion brand aimed at women in their 20s."

[0482] This allows users to quickly generate consistent and accurate ad copy and effectively develop advertising campaigns. In this way, the system of the present invention significantly improves the efficiency and accuracy of ad generation.

[0483] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0484] Step 1:

[0485] A user accesses the system and enters the API key and data path.

[0486] Input: API key, basic data path, corporate data path

[0487] Output: An initialized CustomAIModule instance.

[0488] How it works: A user logs into the system from a smartphone or PC and enters their OpenAI API key and data path. The server receives this information and initializes the instance.

[0489] Step 2:

[0490] The server reads the data from the data storage means.

[0491] Input: Basic Data Path, Corporate Data Path

[0492] Output: Data frame

[0493] What it does: The server uses the Pandas library to read the CSV file from the specified path and convert the base and company data into a dataframe.

[0494] Step 3:

[0495] The server executes the preprocessing means to process the data.

[0496] Input: DataFrame

[0497] Output: Preprocessed dataset

[0498] What it does: The server preprocesses the data frame, handling missing values, removing unnecessary items, and filtering the data, resulting in a clean, unified dataset.

[0499] Step 4:

[0500] The server uses the preprocessed data to perform model training.

[0501] Input: Preprocessed dataset

[0502] Output: A trained generative AI model

[0503] How it works: The server uses the Transformers library and TensorFlow to train a generative AI model using a preprocessed dataset. After training is complete, a highly accurate generative AI model is created.

[0504] Step 5:

[0505] The user enters a prompt statement.

[0506] Input: prompt statement

[0507] Output: Tokenized prompt

[0508] How it works: The user inputs a specific prompt, such as "Please write an ad copy for a new dress from a fashion brand aimed at women in their 20s," from their smartphone or computer. The server receives this prompt and performs a tokenization process.

[0509] Step 6:

[0510] The server generates the advertisement copy using the generation means.

[0511] Input: Tokenized prompt

[0512] Output: Generated ad text

[0513] How it works: The server feeds the tokenized prompts to a trained generative AI model to generate optimal ad copy, which is then displayed to the user.

[0514] Step 7:

[0515] The user reviews and uses the generated ad copy.

[0516] Input: Generated ad text

[0517] Output: Available ad text

[0518] How it works: Users review the ad copy provided by the system, modify or edit it as needed, and use it as the final ad copy, allowing for efficient and consistent ad campaigns.

[0519] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0520] The system of the present invention includes a data storage means, a pre-processing means, a model training means, a generation means, and an emotion engine.

[0521] First, the user configures the system by entering the OpenAI API key, the base data path, and the company's proprietary data path, and then initializes an instance of the CustomAIModule class. This gets the system ready to run.

[0522] The server accesses the data storage means that holds the basic data and the company's proprietary data, and reads the CSV files from each data path. The pre-processing means performs data cleansing and filtering on the data to generate an integrated data set.

[0523] The server then executes a model training process using the preprocessed data, which includes the basic data, the company-specific data, and the emotion information data, to create a highly accurate text generation model with emotion recognition capabilities.

[0524] After the model is trained, users can input specific prompts into the system, such as "Please give me some advice on how to respond to a customer complaint." The server tokenizes the prompt and provides it to the emotion engine, which analyzes the prompt content and classifies the user's emotion.

[0525] The server then feeds the emotion information classified by the emotion engine to a trained text generation model, and uses a generation means to generate an optimal response text, which includes an appropriate response according to the emotion.

[0526] As a concrete example, consider a customer support scenario. The user (customer support representative) sets the AI's API key, basic data path, and company-specific data (e.g., past customer interaction history). The server reads and preprocesses the basic data, company-specific data, and past customer interaction data to create a training dataset.

[0527] The server trains the model using the training data and displays a message indicating completion of training. When a customer support representative inputs "What should I do if a customer is angry about a product?" into the generative AI, the server analyzes the input prompt with an emotion engine and classifies the emotion. The server generates a response text according to the emotion and provides it to the representative.

[0528] In this way, the system of the present invention is able to recognize the user's emotions and quickly provide highly accurate and reliable information in response to the emotions.

[0529] The processing flow will be explained below.

[0530] Step 1:

[0531] The user sets the OpenAI API key, basic data path, and the company's own data path, and initializes an instance of the CustomAIModule class.

[0532] Step 2:

[0533] The server accesses a data storage means that holds basic data and company-specific data, and reads CSV files from each data path.

[0534] Step 3:

[0535] The server uses preprocessing tools to perform data cleansing on the loaded data, including handling missing values ​​and removing unnecessary columns.

[0536] Step 4:

[0537] The server combines the base data and the company's proprietary data into one integrated data set, so that all data is in a unified format.

[0538] Step 5:

[0539] The server executes a model training means to train a text generation model using the preprocessed data, which involves extracting and tokenizing text data from the dataset.

[0540] Step 6:

[0541] The server adds emotion information to the training data and creates a new training set that includes information corresponding to emotions.

[0542] Step 7:

[0543] The server begins training the text generation model using the tokenized text data and sentiment information. The training process is configured to span multiple epochs, during which the model learns the association between text features and sentiment.

[0544] Step 8:

[0545] The server confirms that training is complete and displays a message to the user saying "Model training is complete."

[0546] Step 9:

[0547] The user enters a specific prompt into the system (e.g., "What should I do if a customer is upset about my product?").

[0548] Step 10:

[0549] The server tokenizes the input prompt and feeds it to the emotion engine, which analyzes the user's input and classifies the emotion.

[0550] Step 11:

[0551] The server supplies the emotion information classified by the emotion engine to a trained text generation model, and generates an optimal response text using a generation means.

[0552] Step 12:

[0553] The server decodes the generated response text and presents it to the user.

[0554] Step 13:

[0555] The user checks the response text provided and uses it in their work as needed.

[0556] Example 2

[0557] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0558] Currently, there is a lack of systems that can accurately recognize customer emotions and quickly provide appropriate responses. This results in a problem of reduced efficiency of human resources and time in customer service operations in various fields, where efficient and effective support is required. In light of this situation, there is a need to develop a system that can achieve highly accurate emotion recognition and response generation.

[0559] 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.

[0560] In this invention, the server includes a user setting means, a data storage means, a preprocessing means, a model training means, a sentiment analysis means, and a text generation means, which enable the server to recognize the user's sentiment and quickly provide highly accurate and reliable information in response to the sentiment.

[0561] The "user setting means" is a function that allows the user to input setting information required to start the system operation.

[0562] "Data storage means" refers to a storage device or system for holding basic data, company-specific data, and sentiment information data.

[0563] The "preprocessing means" is a function that performs missing value complementation, correction of inconsistent data, and filtering of the acquired data to generate a clean dataset.

[0564] The "model training means" is a function that uses preprocessed data to train a machine learning model and create a highly accurate text generation model.

[0565] The "emotion analysis means" is a function that analyzes the user's emotions from the input text data and assigns appropriate emotion labels.

[0566] The "text generation means" is a function that generates optimal response text based on the emotion data obtained by the emotion analysis means.

[0567] The system of the present invention realizes highly accurate text generation by allowing the user to configure the system and input specific prompt sentences. Specifically, the system includes the following means: user configuration means, data storage means, preprocessing means, model training means, sentiment analysis means, and text generation means.

[0568] First, the user enters the necessary configuration information to set up the system, including the API key, the base data path, and the company's own data path. This initializes an instance of the CustomAIModule class, and the system is ready to run.

[0569] The server accesses the data storage means and reads the basic data, company-specific data, and emotion information data in CSV format. The data storage means refers to a storage device for storing these data.

[0570] Next, the server preprocesses the data using a preprocessing method, which performs missing value imputation, inconsistent data correction, duplicate data removal, and data filtering to generate a clean dataset. Specifically, the server performs data cleansing and filtering on the loaded data.

[0571] The server then uses a model training means to train a machine learning model based on the preprocessed data. The training data includes basic data, company-specific data, and sentiment information data. The training process is repeated to build a highly accurate text generation model. When training is complete, the server displays a message indicating that training is complete.

[0572] The user can input a specific prompt to the system. For example, "Please give me some advice on how to respond to a customer complaint." The server tokenizes this input and provides it to the sentiment analyzer. The sentiment analyzer analyzes the prompt content and classifies the user's sentiment.

[0573] Next, the server supplies the emotion information obtained by the emotion analysis means to the trained text generation model, and uses the text generation means to generate an optimal response text, which includes an appropriate response according to the emotion.

[0574] A concrete example of this would be a customer support scenario. The user (customer support representative) sets an API key, a basic data path, and the company's proprietary data (e.g., past customer interaction history). The server loads and preprocesses the basic data, the company's proprietary data, and past customer interaction data to create a training dataset. The server then trains the model using the training data, and a message indicating training completion is displayed. When the customer support representative inputs the following into the generative AI: "What should I do if a customer is upset about a product?", the server analyzes this input using the emotion engine and assigns an appropriate emotion label. Finally, the server generates the optimal response text based on the emotion and provides it to the representative.

[0575] In this way, the system of the present invention is able to recognize the user's emotions and quickly provide highly accurate and reliable information in response to the emotions.

[0576] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0577] Step 1:

[0578] The user configures the system.

[0579] Input: OpenAI API key, basic data path, company's proprietary data path

[0580] Operation: The user enters this information on the screen and sends it to the system. An instance of the CustomAIModule class is initialized and the system is ready to operate.

[0581] Output: System completed initialization

[0582] Step 2:

[0583] The server accesses the data storage means and reads the data.

[0584] Input: Basic data path, company's proprietary data path

[0585] Operation: Accesses the data storage means and reads the CSV files from each data path.

[0586] Output: Raw dataset of loaded basic and company data

[0587] Step 3:

[0588] The server pre-processes the data using the pre-processing means.

[0589] Input: Raw dataset

[0590] What it does: It performs data cleansing and filtering on the data it loads, including imputing missing values, correcting inconsistent data, removing duplicates, and filtering data.

[0591] Output: A clean dataset

[0592] Step 4:

[0593] The server trains the model using the model training means.

[0594] Input: Clean dataset, basic data, company data, sentiment data

[0595] How it works: A machine learning algorithm is used to train a model. The training process is repeated to build a highly accurate text generation model.

[0596] Output: A fully trained, highly accurate text generation model

[0597] Step 5:

[0598] The user enters a specific prompt.

[0599] Input: Prompt text (e.g., "Please advise how to handle a customer complaint.")

[0600] Action: Sends the entered prompt to the system.

[0601] Output: The prompt text entered into the system

[0602] Step 6:

[0603] The server parses the prompt.

[0604] Input: prompt statement

[0605] How it works: The prompt sentence is tokenized and fed to the sentiment analyzer, which analyzes the prompt content and classifies the user's sentiment.

[0606] Output: Classified emotion data

[0607] Step 7:

[0608] The server generates a response text based on the emotion information.

[0609] Input: Classified emotion data, trained text generation model

[0610] Operation: Generate optimal response text based on emotional information obtained by the emotion analysis means. Generate highly accurate text using the generation means.

[0611] Output: Optimal response text depending on the sentiment

[0612] Step 8:

[0613] The server provides the response text to the user.

[0614] Input: Response text

[0615] Behavior: The server generates a response text and displays it on the user's screen.

[0616] Output: The response text provided to the user

[0617] (Application example 2)

[0618] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0619] Conventional customer support systems have difficulty in recognizing and responding to users' emotions appropriately, making it difficult to increase user satisfaction, especially on online shopping sites. Returning a uniform response without considering emotions can further increase user dissatisfaction.

[0620] 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.

[0621] In this invention, the server includes a data storage means, a preprocessing means, a model training means, a user response generation means, and an emotion recognition means, which enable the server to analyze prompts input by a user, appropriately recognize the emotion, and generate a highly accurate response accordingly.

[0622] The "data storage means" is a storage device for storing data entered by users, academic knowledge data, multilingual data, legal data, general knowledge data, and company-specific data.

[0623] The "preprocessing means" is a device or software that performs cleansing such as processing missing values, deleting unnecessary columns, and filtering data on the data acquired from the data storage means, and integrates the data.

[0624] A "model training means" is a device or software for training a text generation model using preprocessed data.

[0625] A "generator" is a device or software that uses a trained model to generate response text based on a user's input.

[0626] An "emotion recognition means" is a device or software that analyzes the prompt entered by the user and classifies the emotion from its content.

[0627] The "user response generation means" is a device or software that generates an optimal response text based on the emotion information classified by the emotion recognition means.

[0628] A "prompt" is a question or request that a user enters into a system.

[0629] A "training dataset" is a collection of data that has been preprocessed for training a model.

[0630] "Emotion information" is information that indicates the type and intensity of the emotion contained in the user's prompt, as analyzed by the emotion recognition means.

[0631] A "text generation model" is a model that learns from a training dataset to generate appropriate text responses based on user prompts.

[0632] The system of the present invention includes a data storage means, a preprocessing means, a model training means, a generation means, an emotion recognition means, and a user response generation means. This system is specifically implemented as follows.

[0633] The system of the present invention first stores basic data entered by users, company-specific data, and emotion information data in a data storage means. These data are often saved as CSV files.

[0634] Next, the preprocessing means reads the data from the data storage means, handles missing values, removes unnecessary columns, and filters the data to generate a unified dataset, which provides consistent data for subsequent training processes.

[0635] The model training method uses preprocessed data to train a generative AI model. This training includes basic data, company-specific data, and sentiment data. The program is typically implemented using an existing language model API, such as the OpenAI API.

[0636] Once training is complete, the user response generator operates and accepts prompts from the user. For example, if the user inputs "A customer is angry about the product. What should I do?", the system analyzes the prompt with the emotion recognition unit and classifies the user's emotion.

[0637] The emotion information recognized by the emotion recognition means is supplied to the generation means, which generates an appropriate response text. The generated response is in line with the user's emotion and is provided with high accuracy and speed.

[0638] Specifically, when a user uses a customer support application on their smartphone and inputs, "My order hasn't arrived yet. What's going on?", the system first analyzes the input using emotion recognition and recognizes the emotion of "dissatisfaction (anger)." It then uses a generative AI model to generate an appropriate response text, providing the user with a support message such as, "We apologize for the delay in your order. We will check the current situation and get back to you as soon as possible."

[0639] Therefore, the system of the present invention can solve the problems that conventional systems have had and provide highly accurate customer support that responds to the user's emotions.

[0640] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0641] Step 1:

[0642] A user inputs a prompt sentence through a smartphone application, for example, "A customer is angry about the product. What should I do?" This input is sent to the server in text format.

[0643] Step 2:

[0644] The server reads the necessary data from the data storage means, which includes basic data, company-specific data, and sentiment data. The data is stored as CSV files, and the server parses these files.

[0645] Input: CSV file stored in your data storage means

[0646] Output: Raw data (basic data, company-specific data, sentiment data)

[0647] Step 3:

[0648] Preprocessing procedures are used on the raw data to handle missing values, remove unnecessary columns, and filter the data, creating a consistent, integrated dataset.

[0649] Input: Raw data read

[0650] Output: A consolidated dataset

[0651] Step 4:

[0652] The server uses a model training method to train a generative AI model using the pre-processed dataset, which is often done using an external language model API such as the OpenAI API.

[0653] Input: Preprocessed dataset

[0654] Output: A trained generative AI model

[0655] Step 5:

[0656] The server receives the prompt sentence entered by the user and analyzes it with the emotion recognition means, which analyzes the text of the prompt sentence and classifies the user's emotion.

[0657] Input: User prompt text

[0658] Output: Classified emotion information

[0659] Step 6:

[0660] The classified emotion information is supplied to the server's generation means, which generates a response text that matches the emotion. The generation means uses a trained generative AI model to generate the optimal response sentence.

[0661] Input: Classified emotion information, user prompt

[0662] Output: The generated response text

[0663] Step 7:

[0664] The response text generated by the response generating means is sent to the user terminal, and the user can view the generated response through a smartphone application.

[0665] Input: Generated response text

[0666] Output: The response message that is displayed to the user

[0667] The above steps enable highly accurate customer support that takes into account the user's emotions.

[0668] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0669] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0670] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0671] [Third embodiment]

[0672] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0673] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0674] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0675] 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.

[0676] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0677] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0678] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0679] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0680] 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.

[0681] 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.

[0682] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0683] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0684] The system of the present invention includes a data storage means, a pre-processing means, a model training means, and a generation means.

[0685] First, the user configures the system by entering the OpenAI API key, the base data path, and the company's proprietary data path, which then initializes an instance of the CustomAIModule class.

[0686] The server has a data storage means for storing basic data and company-specific data, reads CSV files from each data path, and converts them into data frames. The pre-processing means performs data cleansing and filtering on the read data to generate an integrated dataset.

[0687] The server then runs a model training procedure using the pre-processed data, which includes both baseline data and company-specific data, to create a more accurate text generation model.

[0688] After the model has been trained, a user can input a specific prompt to the system, such as "Please give me some advice on my company's sales strategy." The server tokenizes the prompt and provides it to the trained model. The generator generates an optimal response based on the input prompt and provides the response to the user.

[0689] As a concrete example, consider a usage scenario in a hospital. The user (hospital system administrator) sets the AI's API key, basic data path, and the hospital's proprietary data (e.g., patient treatment data). The server reads the basic data and the hospital's proprietary data, preprocesses them, and creates a training dataset.

[0690] The server trains the GPT-2 model using the training data and displays a message indicating completion of training. When a doctor at the hospital inputs "Please give me advice on the latest treatments" into the generative AI, the server generates information on appropriate treatments from the trained model based on the input prompt and provides it to the doctor.

[0691] In this way, the system of the present invention can quickly provide highly reliable information and realize highly accurate output that is directly linked to specific business operations.

[0692] The processing flow will be explained below.

[0693] Step 1:

[0694] The user sets the OpenAI API key, basic data path, and the company's own data path, and initializes an instance of the CustomAIModule class.

[0695] Step 2:

[0696] The server accesses a data storage means that holds basic data and company-specific data, and reads CSV files from each data path.

[0697] Step 3:

[0698] The server uses preprocessing tools to cleanse the ingested data by handling missing values ​​and removing unnecessary columns.

[0699] Step 4:

[0700] The server combines the base data and the company's proprietary data into one integrated data set, so that all data is in a unified format.

[0701] Step 5:

[0702] The server executes a model training means to train a text generation model using the preprocessed data, which involves extracting and tokenizing text data from the dataset.

[0703] Step 6:

[0704] The server begins training the text generation model using the tokenized text data. The training process is set up over multiple epochs, during which the model learns the features of the text.

[0705] Step 7:

[0706] The server confirms that training is complete and displays a message to the user saying "Model training is complete."

[0707] Step 8:

[0708] The user inputs a specific prompt (e.g., "Please give me some advice on our company's sales strategy") to the system.

[0709] Step 9:

[0710] The server tokenizes the input prompt and feeds it into a trained text generation model.

[0711] Step 10:

[0712] The server uses the trained model to generate text responses based on the prompts, ensuring accurate and relevant responses.

[0713] Step 11:

[0714] The server decodes the generated response text and presents it to the user.

[0715] Step 12:

[0716] The user checks the response text provided and uses it in their work as needed.

[0717] Example 1

[0718] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0719] In information generation systems using conventional generative AI models, it has been difficult to train highly accurate models that combine reliable data with diverse data sources, and it has also been challenging to quickly and appropriately provide optimal responses to prompts from users. In particular, when customized information is required for each company or industry, a method for efficiently handling training data that matches that needs has been required.

[0720] 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.

[0721] In this invention, the server includes a data storage unit, a preprocessing unit, a model training unit, a generation unit, a user setting input unit, a prompt input unit, a response generation unit, and a presentation unit. This allows for the effective integration of a diverse set of reliable data to train a highly accurate generative AI model customized for each company or industry. This allows for the provision of prompt and appropriate responses to specific prompts from users.

[0722] "Data storage means" refers to a means for storing various data including basic data and company-specific data provided by users.

[0723] The "pre-processing means" is a means for performing pre-processing such as cleansing, filtering, and integration on the data read from the data storage means.

[0724] "Model training means" means a means for training a generative AI model using preprocessed data.

[0725] "Generation means" means a means for generating an optimal response to a prompt input by a user using a trained generative AI model.

[0726] The "user setting input means" is a means by which a user inputs an API key, a basic data path, and a company-specific data path into the system.

[0727] A "prompt input means" is a means by which a user inputs specific questions or requests into the system.

[0728] A "response generation means" is a means for tokenizing an input prompt and generating a response using a trained generative AI model.

[0729] A "presentation means" is a means for displaying the generated response to the user.

[0730] The system of the present invention includes a data storage means, a preprocessing means, a model training means, a generation means, a user setting input means, a prompt input means, a response generation means, and a presentation means.

[0731] First, the user configures the system by entering an API key, a base data path, and a company-specific data path. This configuration initializes an instance of the CustomAIModule class. The data path entered by the user contains customizable data for each company.

[0732] Next, the server stores the basic data and company-specific data through a data storage means. The server reads the CSV file from the data path specified by the user and converts it into a data frame using the Pandas library.

[0733] A preprocessing tool then performs data cleansing and filtering on the data frames, including handling missing values, removing unnecessary columns, and filtering the data to produce a single unified dataset.

[0734] The server then runs a model training procedure using the combined dataset. The training dataset includes both baseline and company-specific data, resulting in a more accurate generative AI model. This training utilizes deep learning frameworks such as TensorFlow and PyTorch. Training can take anywhere from several hours to several days.

[0735] Once the model is trained, a user can enter a specific prompt through the prompt input means, for example, "Please give me some advice on my company's sales strategy." The server tokenizes this prompt and feeds the preprocessed tokens to the generative AI model.

[0736] The response generation means uses the trained model to generate an optimal response based on the input prompt. This response is provided to the user through the presentation means. For example, in response to the prompt "Please give me some advice on my company's sales strategy," the trained model generates advice on an appropriate sales strategy and provides it to the user.

[0737] As a concrete example, consider a usage scenario in a hospital. The user (hospital system administrator) sets the API key, basic data path, and hospital-specific data path (e.g., patient treatment data). The server reads the basic data and the hospital-specific data, preprocesses them, and creates an integrated dataset. The server trains the generative AI model using the training data and displays a message indicating training completion. When a doctor at the hospital inputs to the generative AI, "Please give me some advice on the latest treatments," the server generates information on the optimal treatment based on the input prompt and provides it to the doctor.

[0738] In this way, the system of the present invention can quickly provide highly reliable information and realize highly accurate output that is directly linked to specific business operations.

[0739] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0740] Step 1:

[0741] User-configured system

[0742] Input: API key, basic data path, company's own data path

[0743] Description: The user enters the API key, basic data path, and company-specific data path in the system settings screen. This setting registers the necessary information in the system and initializes an instance of the CustomAIModule class.

[0744] Output: An initialized instance of the CustomAIModule class.

[0745] Step 2:

[0746] Data loading

[0747] Input: User specified data path

[0748] Description: The server uses the data storage means to read the CSV file from the basic data path and company-specific data path entered by the user. At this time, the CSV file is converted to a data frame using the Pandas library and stored in memory.

[0749] Output: Basic and company-specific data in a data frame format

[0750] Step 3:

[0751] Data Preprocessing

[0752] Input: Basic data and company-specific data in a data frame format

[0753] Description: The preprocessing method performs data cleansing and filtering on the data frame, specifically handling missing values, removing unnecessary columns, and filtering the data to produce a single unified dataset.

[0754] Output: A consolidated dataset

[0755] Step 4:

[0756] Model Training

[0757] Input: Integrated dataset

[0758] Description: The server runs a model training procedure using the combined dataset. The training dataset includes both baseline and company-specific data, resulting in a more accurate generative AI model. This training utilizes deep learning frameworks such as TensorFlow and PyTorch. Training can take anywhere from several hours to several days.

[0759] Output: A trained generative AI model

[0760] Step 5:

[0761] Prompt Input

[0762] Input: A specific prompt from the user (e.g., "Please give me some advice on our company's sales strategy.")

[0763] Description: The user enters a specific prompt into the system, which contains information needed to generate a future response.

[0764] Output: The prompt entered

[0765] Step 6:

[0766] Prompt Tokenization

[0767] Input: Entered prompt

[0768] Description: The server tokenizes the input prompt by splitting it into words or sub-words and converting each into a token, often using Hugging Face's Transformers library.

[0769] Output: Tokenized prompt

[0770] Step 7:

[0771] Response Generation

[0772] Input: Tokenized prompt, trained generative AI model

[0773] Description: The response generator uses a trained model to generate optimal responses to tokenized prompts.

[0774] Output: The generated response

[0775] Step 8:

[0776] Response suggestion

[0777] Input: Generated response

[0778] Description: The generated response is provided to the user through a presentation method. The response is displayed on the user's device and used as necessary information.

[0779] Output: The response displayed on the user's terminal

[0780] (Application example 1)

[0781] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0782] In recent years, customer targeting has become more sophisticated in the advertising industry, creating a demand for efficient ad generation that utilizes a variety of data. However, traditional ad generation methods require manual copywriting and targeting, which is time-consuming and labor-intensive, and it is difficult to generate consistent ad copy. To solve this problem, a system is needed that utilizes corporate and market data and automatically generates accurate ad copy based on user prompts.

[0783] 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.

[0784] In this invention, the server includes a data storage means, a preprocessing means, a model training means, and an advertising copy generation means, which preprocesses company data and market data and enables the automatic generation of consistent and highly accurate advertising copy.

[0785] "Data storage means" means means for holding basic data, company-specific data, academic knowledge data, multilingual data, legal data, general knowledge data and market data.

[0786] The "preprocessing means" is a means for reading data from the data storage means, processing missing values, deleting unnecessary columns, and filtering the data.

[0787] A "model training means" is a means for training a natural language processing model using preprocessed data and constructing a generative AI model.

[0788] "Generating means" means using a trained generative AI model to generate an optimal response based on prompts from a user.

[0789] The "advertising copy generation means" is a means for automatically generating consistent and highly accurate advertising copy based on company data and market data in response to user prompts.

[0790] "Company Data" means proprietary data related to a specific company that is used for targeting and personalization in advertising generation.

[0791] "Market Data" means data related to a particular market or industry and used for trend analysis and targeting.

[0792] This invention is a system for automatically generating advertising copy, comprising a data storage means, a preprocessing means, a model training means, a generation means, and an advertising copy generation means, which utilizes company data and market data to generate accurate advertising copy based on specific prompts from a user.

[0793] Data Storage Means

[0794] The server includes data storage means for holding basic data, company-specific data, academic knowledge data, multilingual data, legal data, general knowledge data, and market data. This data storage means includes hard disk drives and database systems required for data capture and management.

[0795] Pretreatment means

[0796] The server has a preprocessing means for reading data from the data storage means, processing missing values, deleting unnecessary items, filtering data, etc. For example, the server uses a data processing tool such as the Pandas library to efficiently preprocess the data.

[0797] Model training methods

[0798] Using the preprocessed data, a model training method builds a generative AI model, which uses a machine learning framework such as Python's Transformers library or TensorFlow for training.

[0799] generation means

[0800] The generator uses a trained generative AI model to generate an optimal response based on the user's prompts. Specifically, it uses the OpenAI API to invoke a natural language processing model and generate text based on the user's request.

[0801] Ad copy generation method

[0802] The advertising copy generation means automatically generates targeted advertising copy based on a prompt text provided by a user, utilizing company data and market data.

[0803] For example, if a user inputs a prompt such as "Please write an ad copy for a new dress from a fashion brand aimed at women in their 20s" on their smartphone or computer, the server tokenizes the prompt and feeds it to the trained model. The generated ad copy is then displayed.

[0804] For example, consider the following prompt:

[0805] "Please write an ad copy for a new dress from a fashion brand aimed at women in their 20s."

[0806] This allows users to quickly generate consistent and accurate ad copy and effectively develop advertising campaigns. In this way, the system of the present invention significantly improves the efficiency and accuracy of ad generation.

[0807] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0808] Step 1:

[0809] A user accesses the system and enters the API key and data path.

[0810] Input: API key, basic data path, corporate data path

[0811] Output: An initialized CustomAIModule instance.

[0812] How it works: A user logs into the system from a smartphone or PC and enters their OpenAI API key and data path. The server receives this information and initializes the instance.

[0813] Step 2:

[0814] The server reads the data from the data storage means.

[0815] Input: Basic Data Path, Corporate Data Path

[0816] Output: Data frame

[0817] What it does: The server uses the Pandas library to read the CSV file from the specified path and convert the base and company data into a dataframe.

[0818] Step 3:

[0819] The server executes the preprocessing means to process the data.

[0820] Input: DataFrame

[0821] Output: Preprocessed dataset

[0822] What it does: The server preprocesses the data frame, handling missing values, removing unnecessary items, and filtering the data, resulting in a clean, unified dataset.

[0823] Step 4:

[0824] The server uses the preprocessed data to perform model training.

[0825] Input: Preprocessed dataset

[0826] Output: A trained generative AI model

[0827] How it works: The server uses the Transformers library and TensorFlow to train a generative AI model using a preprocessed dataset. After training is complete, a highly accurate generative AI model is created.

[0828] Step 5:

[0829] The user enters a prompt statement.

[0830] Input: prompt statement

[0831] Output: Tokenized prompt

[0832] How it works: The user inputs a specific prompt, such as "Please write an ad copy for a new dress from a fashion brand aimed at women in their 20s," from their smartphone or computer. The server receives this prompt and performs a tokenization process.

[0833] Step 6:

[0834] The server generates the advertisement copy using the generation means.

[0835] Input: Tokenized prompt

[0836] Output: Generated ad text

[0837] How it works: The server feeds the tokenized prompts to a trained generative AI model to generate optimal ad copy, which is then displayed to the user.

[0838] Step 7:

[0839] The user reviews and uses the generated ad copy.

[0840] Input: Generated ad text

[0841] Output: Available ad text

[0842] How it works: Users review the ad copy provided by the system, modify or edit it as needed, and use it as the final ad copy, allowing for efficient and consistent ad campaigns.

[0843] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0844] The system of the present invention includes a data storage means, a pre-processing means, a model training means, a generation means, and an emotion engine.

[0845] First, the user configures the system by entering the OpenAI API key, the base data path, and the company's proprietary data path, and then initializes an instance of the CustomAIModule class. This gets the system ready to run.

[0846] The server accesses the data storage means that holds the basic data and the company's proprietary data, and reads the CSV files from each data path. The pre-processing means performs data cleansing and filtering on the data to generate an integrated data set.

[0847] The server then executes a model training process using the preprocessed data, which includes the basic data, the company-specific data, and the emotion information data, to create a highly accurate text generation model with emotion recognition capabilities.

[0848] After the model is trained, users can input specific prompts into the system, such as "Please give me some advice on how to respond to a customer complaint." The server tokenizes the prompt and provides it to the emotion engine, which analyzes the prompt content and classifies the user's emotion.

[0849] The server then feeds the emotion information classified by the emotion engine to a trained text generation model, and uses a generation means to generate an optimal response text, which includes an appropriate response according to the emotion.

[0850] As a concrete example, consider a customer support scenario. The user (customer support representative) sets the AI's API key, basic data path, and company-specific data (e.g., past customer interaction history). The server reads and preprocesses the basic data, company-specific data, and past customer interaction data to create a training dataset.

[0851] The server trains the model using the training data and displays a message indicating completion of training. When a customer support representative inputs "What should I do if a customer is angry about a product?" into the generative AI, the server analyzes the input prompt with an emotion engine and classifies the emotion. The server generates a response text according to the emotion and provides it to the representative.

[0852] In this way, the system of the present invention is able to recognize the user's emotions and quickly provide highly accurate and reliable information in response to the emotions.

[0853] The processing flow will be explained below.

[0854] Step 1:

[0855] The user sets the OpenAI API key, basic data path, and the company's own data path, and initializes an instance of the CustomAIModule class.

[0856] Step 2:

[0857] The server accesses a data storage means that holds basic data and company-specific data, and reads CSV files from each data path.

[0858] Step 3:

[0859] The server uses preprocessing tools to perform data cleansing on the loaded data, including handling missing values ​​and removing unnecessary columns.

[0860] Step 4:

[0861] The server combines the base data and the company's proprietary data into one integrated data set, so that all data is in a unified format.

[0862] Step 5:

[0863] The server executes a model training means to train a text generation model using the preprocessed data, which involves extracting and tokenizing text data from the dataset.

[0864] Step 6:

[0865] The server adds emotion information to the training data and creates a new training set that includes information corresponding to emotions.

[0866] Step 7:

[0867] The server begins training the text generation model using the tokenized text data and sentiment information. The training process is configured to span multiple epochs, during which the model learns the association between text features and sentiment.

[0868] Step 8:

[0869] The server confirms that training is complete and displays a message to the user saying "Model training is complete."

[0870] Step 9:

[0871] The user enters a specific prompt into the system (e.g., "What should I do if a customer is upset about my product?").

[0872] Step 10:

[0873] The server tokenizes the input prompt and feeds it to the emotion engine, which analyzes the user's input and classifies the emotion.

[0874] Step 11:

[0875] The server supplies the emotion information classified by the emotion engine to a trained text generation model, and generates an optimal response text using a generation means.

[0876] Step 12:

[0877] The server decodes the generated response text and presents it to the user.

[0878] Step 13:

[0879] The user checks the response text provided and uses it in their work as needed.

[0880] Example 2

[0881] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0882] Currently, there is a lack of systems that can accurately recognize customer emotions and quickly provide appropriate responses. This results in a problem of reduced efficiency of human resources and time in customer service operations in various fields, where efficient and effective support is required. In light of this situation, there is a need to develop a system that can achieve highly accurate emotion recognition and response generation.

[0883] 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.

[0884] In this invention, the server includes a user setting means, a data storage means, a preprocessing means, a model training means, a sentiment analysis means, and a text generation means, which enable the server to recognize the user's sentiment and quickly provide highly accurate and reliable information in response to the sentiment.

[0885] The "user setting means" is a function that allows the user to input setting information required to start the system operation.

[0886] "Data storage means" refers to a storage device or system for holding basic data, company-specific data, and sentiment information data.

[0887] The "preprocessing means" is a function that performs missing value complementation, correction of inconsistent data, and filtering of the acquired data to generate a clean dataset.

[0888] The "model training means" is a function that uses preprocessed data to train a machine learning model and create a highly accurate text generation model.

[0889] The "emotion analysis means" is a function that analyzes the user's emotions from the input text data and assigns appropriate emotion labels.

[0890] The "text generation means" is a function that generates optimal response text based on the emotion data obtained by the emotion analysis means.

[0891] The system of the present invention realizes highly accurate text generation by allowing the user to configure the system and input specific prompt sentences. Specifically, the system includes the following means: user configuration means, data storage means, preprocessing means, model training means, sentiment analysis means, and text generation means.

[0892] First, the user enters the necessary configuration information to set up the system, including the API key, the base data path, and the company's own data path. This initializes an instance of the CustomAIModule class, and the system is ready to run.

[0893] The server accesses the data storage means and reads the basic data, company-specific data, and emotion information data in CSV format. The data storage means refers to a storage device for storing these data.

[0894] Next, the server preprocesses the data using a preprocessing method, which performs missing value imputation, inconsistent data correction, duplicate data removal, and data filtering to generate a clean dataset. Specifically, the server performs data cleansing and filtering on the loaded data.

[0895] The server then uses a model training means to train a machine learning model based on the preprocessed data. The training data includes basic data, company-specific data, and sentiment information data. The training process is repeated to build a highly accurate text generation model. When training is complete, the server displays a message indicating that training is complete.

[0896] The user can input a specific prompt to the system. For example, "Please give me some advice on how to respond to a customer complaint." The server tokenizes this input and provides it to the sentiment analyzer. The sentiment analyzer analyzes the prompt content and classifies the user's sentiment.

[0897] Next, the server supplies the emotion information obtained by the emotion analysis means to the trained text generation model, and uses the text generation means to generate an optimal response text, which includes an appropriate response according to the emotion.

[0898] A concrete example of this would be a customer support scenario. The user (customer support representative) sets an API key, a basic data path, and the company's proprietary data (e.g., past customer interaction history). The server loads and preprocesses the basic data, the company's proprietary data, and past customer interaction data to create a training dataset. The server then trains the model using the training data, and a message indicating training completion is displayed. When the customer support representative inputs the following into the generative AI: "What should I do if a customer is upset about a product?", the server analyzes this input using the emotion engine and assigns an appropriate emotion label. Finally, the server generates the optimal response text based on the emotion and provides it to the representative.

[0899] In this way, the system of the present invention is able to recognize the user's emotions and quickly provide highly accurate and reliable information in response to the emotions.

[0900] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0901] Step 1:

[0902] The user configures the system.

[0903] Input: OpenAI API key, basic data path, company's proprietary data path

[0904] Operation: The user enters this information on the screen and sends it to the system. An instance of the CustomAIModule class is initialized and the system is ready to operate.

[0905] Output: System completed initialization

[0906] Step 2:

[0907] The server accesses the data storage means and reads the data.

[0908] Input: Basic data path, company's proprietary data path

[0909] Operation: Accesses the data storage means and reads the CSV files from each data path.

[0910] Output: Raw dataset of loaded basic and company data

[0911] Step 3:

[0912] The server pre-processes the data using the pre-processing means.

[0913] Input: Raw dataset

[0914] What it does: It performs data cleansing and filtering on the data it loads, including imputing missing values, correcting inconsistent data, removing duplicates, and filtering data.

[0915] Output: A clean dataset

[0916] Step 4:

[0917] The server trains the model using the model training means.

[0918] Input: Clean dataset, basic data, company data, sentiment data

[0919] How it works: A machine learning algorithm is used to train a model. The training process is repeated to build a highly accurate text generation model.

[0920] Output: A fully trained, highly accurate text generation model

[0921] Step 5:

[0922] The user enters a specific prompt.

[0923] Input: Prompt text (e.g., "Please advise how to handle a customer complaint.")

[0924] Action: Sends the entered prompt to the system.

[0925] Output: The prompt text entered into the system

[0926] Step 6:

[0927] The server parses the prompt.

[0928] Input: prompt statement

[0929] How it works: The prompt sentence is tokenized and fed to the sentiment analyzer, which analyzes the prompt content and classifies the user's sentiment.

[0930] Output: Classified emotion data

[0931] Step 7:

[0932] The server generates a response text based on the emotion information.

[0933] Input: Classified emotion data, trained text generation model

[0934] Operation: Generate optimal response text based on emotional information obtained by the emotion analysis means. Generate highly accurate text using the generation means.

[0935] Output: Optimal response text depending on the sentiment

[0936] Step 8:

[0937] The server provides the response text to the user.

[0938] Input: Response text

[0939] Behavior: The server generates a response text and displays it on the user's screen.

[0940] Output: The response text provided to the user

[0941] (Application example 2)

[0942] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0943] Conventional customer support systems have difficulty in recognizing and responding to users' emotions appropriately, making it difficult to increase user satisfaction, especially on online shopping sites. Returning a uniform response without considering emotions can further increase user dissatisfaction.

[0944] 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.

[0945] In this invention, the server includes a data storage means, a preprocessing means, a model training means, a user response generation means, and an emotion recognition means, which enable the server to analyze prompts input by a user, appropriately recognize the emotion, and generate a highly accurate response accordingly.

[0946] The "data storage means" is a storage device for storing data entered by users, academic knowledge data, multilingual data, legal data, general knowledge data, and company-specific data.

[0947] The "preprocessing means" is a device or software that performs cleansing such as processing missing values, deleting unnecessary columns, and filtering data on the data acquired from the data storage means, and integrates the data.

[0948] A "model training means" is a device or software for training a text generation model using preprocessed data.

[0949] A "generator" is a device or software that uses a trained model to generate response text based on a user's input.

[0950] An "emotion recognition means" is a device or software that analyzes the prompt entered by the user and classifies the emotion from its content.

[0951] The "user response generation means" is a device or software that generates an optimal response text based on the emotion information classified by the emotion recognition means.

[0952] A "prompt" is a question or request that a user enters into a system.

[0953] A "training dataset" is a collection of data that has been preprocessed for training a model.

[0954] "Emotion information" is information that indicates the type and intensity of the emotion contained in the user's prompt, as analyzed by the emotion recognition means.

[0955] A "text generation model" is a model that learns from a training dataset to generate appropriate text responses based on user prompts.

[0956] The system of the present invention includes a data storage means, a preprocessing means, a model training means, a generation means, an emotion recognition means, and a user response generation means. This system is specifically implemented as follows.

[0957] The system of the present invention first stores basic data entered by users, company-specific data, and emotion information data in a data storage means. These data are often saved as CSV files.

[0958] Next, the preprocessing means reads the data from the data storage means, handles missing values, removes unnecessary columns, and filters the data to generate a unified dataset, which provides consistent data for subsequent training processes.

[0959] The model training method uses preprocessed data to train a generative AI model. This training includes basic data, company-specific data, and sentiment data. The program is typically implemented using an existing language model API, such as the OpenAI API.

[0960] Once training is complete, the user response generator operates and accepts prompts from the user. For example, if the user inputs "A customer is angry about the product. What should I do?", the system analyzes the prompt with the emotion recognition unit and classifies the user's emotion.

[0961] The emotion information recognized by the emotion recognition means is supplied to the generation means, which generates an appropriate response text. The generated response is in line with the user's emotion and is provided with high accuracy and speed.

[0962] Specifically, when a user uses a customer support application on their smartphone and inputs, "My order hasn't arrived yet. What's going on?", the system first analyzes the input using emotion recognition and recognizes the emotion of "dissatisfaction (anger)." It then uses a generative AI model to generate an appropriate response text, providing the user with a support message such as, "We apologize for the delay in your order. We will check the current situation and get back to you as soon as possible."

[0963] Therefore, the system of the present invention can solve the problems that conventional systems have had and provide highly accurate customer support that responds to the user's emotions.

[0964] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0965] Step 1:

[0966] A user inputs a prompt sentence through a smartphone application, for example, "A customer is angry about the product. What should I do?" This input is sent to the server in text format.

[0967] Step 2:

[0968] The server reads the necessary data from the data storage means, which includes basic data, company-specific data, and sentiment data. The data is stored as CSV files, and the server parses these files.

[0969] Input: CSV file stored in your data storage means

[0970] Output: Raw data (basic data, company-specific data, sentiment data)

[0971] Step 3:

[0972] Preprocessing procedures are used on the raw data to handle missing values, remove unnecessary columns, and filter the data, creating a consistent, integrated dataset.

[0973] Input: Raw data read

[0974] Output: A consolidated dataset

[0975] Step 4:

[0976] The server uses a model training method to train a generative AI model using the pre-processed dataset, which is often done using an external language model API such as the OpenAI API.

[0977] Input: Preprocessed dataset

[0978] Output: A trained generative AI model

[0979] Step 5:

[0980] The server receives the prompt sentence entered by the user and analyzes it with the emotion recognition means, which analyzes the text of the prompt sentence and classifies the user's emotion.

[0981] Input: User prompt text

[0982] Output: Classified emotion information

[0983] Step 6:

[0984] The classified emotion information is supplied to the server's generation means, which generates a response text that matches the emotion. The generation means uses a trained generative AI model to generate the optimal response sentence.

[0985] Input: Classified emotion information, user prompt

[0986] Output: The generated response text

[0987] Step 7:

[0988] The response text generated by the response generating means is sent to the user terminal, and the user can view the generated response through a smartphone application.

[0989] Input: Generated response text

[0990] Output: The response message that is displayed to the user

[0991] The above steps enable highly accurate customer support that takes into account the user's emotions.

[0992] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0993] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0994] 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.

[0995] [Fourth embodiment]

[0996] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[0997] 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.

[0998] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0999] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1000] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1001] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1002] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1003] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1004] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1005] 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.

[1006] 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.

[1007] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1008] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1009] The system of the present invention includes a data storage means, a pre-processing means, a model training means, and a generation means.

[1010] First, the user configures the system by entering the OpenAI API key, the base data path, and the company's proprietary data path, which then initializes an instance of the CustomAIModule class.

[1011] The server has a data storage means for storing basic data and company-specific data, reads CSV files from each data path, and converts them into data frames. The pre-processing means performs data cleansing and filtering on the read data to generate an integrated dataset.

[1012] The server then runs a model training procedure using the pre-processed data, which includes both baseline data and company-specific data, to create a more accurate text generation model.

[1013] After the model has been trained, a user can input a specific prompt to the system, such as "Please give me some advice on my company's sales strategy." The server tokenizes the prompt and provides it to the trained model. The generator generates an optimal response based on the input prompt and provides the response to the user.

[1014] As a concrete example, consider a usage scenario in a hospital. The user (hospital system administrator) sets the AI's API key, basic data path, and the hospital's proprietary data (e.g., patient treatment data). The server reads the basic data and the hospital's proprietary data, preprocesses them, and creates a training dataset.

[1015] The server trains the GPT-2 model using the training data and displays a message indicating completion of training. When a doctor at the hospital inputs "Please give me advice on the latest treatments" into the generative AI, the server generates information on appropriate treatments from the trained model based on the input prompt and provides it to the doctor.

[1016] In this way, the system of the present invention can quickly provide highly reliable information and realize highly accurate output that is directly linked to specific business operations.

[1017] The processing flow will be explained below.

[1018] Step 1:

[1019] The user sets the OpenAI API key, basic data path, and the company's own data path, and initializes an instance of the CustomAIModule class.

[1020] Step 2:

[1021] The server accesses a data storage means that holds basic data and company-specific data, and reads CSV files from each data path.

[1022] Step 3:

[1023] The server uses preprocessing tools to cleanse the ingested data by handling missing values ​​and removing unnecessary columns.

[1024] Step 4:

[1025] The server combines the base data and the company's proprietary data into one integrated data set, so that all data is in a unified format.

[1026] Step 5:

[1027] The server executes a model training means to train a text generation model using the preprocessed data, which involves extracting and tokenizing text data from the dataset.

[1028] Step 6:

[1029] The server begins training the text generation model using the tokenized text data. The training process is set up over multiple epochs, during which the model learns the features of the text.

[1030] Step 7:

[1031] The server confirms that training is complete and displays a message to the user saying "Model training is complete."

[1032] Step 8:

[1033] The user inputs a specific prompt (e.g., "Please give me some advice on our company's sales strategy") to the system.

[1034] Step 9:

[1035] The server tokenizes the input prompt and feeds it into a trained text generation model.

[1036] Step 10:

[1037] The server uses the trained model to generate text responses based on the prompts, ensuring accurate and relevant responses.

[1038] Step 11:

[1039] The server decodes the generated response text and presents it to the user.

[1040] Step 12:

[1041] The user checks the response text provided and uses it in their work as needed.

[1042] Example 1

[1043] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1044] In information generation systems using conventional generative AI models, it has been difficult to train highly accurate models that combine reliable data with diverse data sources, and it has also been challenging to quickly and appropriately provide optimal responses to prompts from users. In particular, when customized information is required for each company or industry, a method for efficiently handling training data that matches that needs has been required.

[1045] 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.

[1046] In this invention, the server includes a data storage unit, a preprocessing unit, a model training unit, a generation unit, a user setting input unit, a prompt input unit, a response generation unit, and a presentation unit. This allows for the effective integration of a diverse set of reliable data to train a highly accurate generative AI model customized for each company or industry. This allows for the provision of prompt and appropriate responses to specific prompts from users.

[1047] "Data storage means" refers to a means for storing various data including basic data and company-specific data provided by users.

[1048] The "pre-processing means" is a means for performing pre-processing such as cleansing, filtering, and integration on the data read from the data storage means.

[1049] "Model training means" means a means for training a generative AI model using preprocessed data.

[1050] "Generation means" means a means for generating an optimal response to a prompt input by a user using a trained generative AI model.

[1051] The "user setting input means" is a means by which a user inputs an API key, a basic data path, and a company-specific data path into the system.

[1052] A "prompt input means" is a means by which a user inputs specific questions or requests into the system.

[1053] A "response generation means" is a means for tokenizing an input prompt and generating a response using a trained generative AI model.

[1054] A "presentation means" is a means for displaying the generated response to the user.

[1055] The system of the present invention includes a data storage means, a preprocessing means, a model training means, a generation means, a user setting input means, a prompt input means, a response generation means, and a presentation means.

[1056] First, the user configures the system by entering an API key, a base data path, and a company-specific data path. This configuration initializes an instance of the CustomAIModule class. The data path entered by the user contains customizable data for each company.

[1057] Next, the server stores the basic data and company-specific data through a data storage means. The server reads the CSV file from the data path specified by the user and converts it into a data frame using the Pandas library.

[1058] A preprocessing tool then performs data cleansing and filtering on the data frames, including handling missing values, removing unnecessary columns, and filtering the data to produce a single unified dataset.

[1059] The server then runs a model training procedure using the combined dataset. The training dataset includes both baseline and company-specific data, resulting in a more accurate generative AI model. This training utilizes deep learning frameworks such as TensorFlow and PyTorch. Training can take anywhere from several hours to several days.

[1060] Once the model is trained, a user can enter a specific prompt through the prompt input means, for example, "Please give me some advice on my company's sales strategy." The server tokenizes this prompt and feeds the preprocessed tokens to the generative AI model.

[1061] The response generation means uses the trained model to generate an optimal response based on the input prompt. This response is provided to the user through the presentation means. For example, in response to the prompt "Please give me some advice on my company's sales strategy," the trained model generates advice on an appropriate sales strategy and provides it to the user.

[1062] As a concrete example, consider a usage scenario in a hospital. The user (hospital system administrator) sets the API key, basic data path, and hospital-specific data path (e.g., patient treatment data). The server reads the basic data and the hospital-specific data, preprocesses them, and creates an integrated dataset. The server trains the generative AI model using the training data and displays a message indicating training completion. When a doctor at the hospital inputs to the generative AI, "Please give me some advice on the latest treatments," the server generates information on the optimal treatment based on the input prompt and provides it to the doctor.

[1063] In this way, the system of the present invention can quickly provide highly reliable information and realize highly accurate output that is directly linked to specific business operations.

[1064] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1065] Step 1:

[1066] User-configured system

[1067] Input: API key, basic data path, company's own data path

[1068] Description: The user enters the API key, basic data path, and company-specific data path in the system settings screen. This setting registers the necessary information in the system and initializes an instance of the CustomAIModule class.

[1069] Output: An initialized instance of the CustomAIModule class.

[1070] Step 2:

[1071] Data loading

[1072] Input: User specified data path

[1073] Description: The server uses the data storage means to read the CSV file from the basic data path and company-specific data path entered by the user. At this time, the CSV file is converted to a data frame using the Pandas library and stored in memory.

[1074] Output: Basic and company-specific data in a data frame format

[1075] Step 3:

[1076] Data Preprocessing

[1077] Input: Basic data and company-specific data in a data frame format

[1078] Description: The preprocessing method performs data cleansing and filtering on the data frame, specifically handling missing values, removing unnecessary columns, and filtering the data to produce a single unified dataset.

[1079] Output: A consolidated dataset

[1080] Step 4:

[1081] Model Training

[1082] Input: Integrated dataset

[1083] Description: The server runs a model training procedure using the combined dataset. The training dataset includes both baseline and company-specific data, resulting in a more accurate generative AI model. This training utilizes deep learning frameworks such as TensorFlow and PyTorch. Training can take anywhere from several hours to several days.

[1084] Output: A trained generative AI model

[1085] Step 5:

[1086] Prompt Input

[1087] Input: A specific prompt from the user (e.g., "Please give me some advice on our company's sales strategy.")

[1088] Description: The user enters a specific prompt into the system, which contains information needed to generate a future response.

[1089] Output: The prompt entered

[1090] Step 6:

[1091] Prompt Tokenization

[1092] Input: Entered prompt

[1093] Description: The server tokenizes the input prompt by splitting it into words or sub-words and converting each into a token, often using Hugging Face's Transformers library.

[1094] Output: Tokenized prompt

[1095] Step 7:

[1096] Response Generation

[1097] Input: Tokenized prompt, trained generative AI model

[1098] Description: The response generator uses a trained model to generate optimal responses to tokenized prompts.

[1099] Output: The generated response

[1100] Step 8:

[1101] Response suggestion

[1102] Input: Generated response

[1103] Description: The generated response is provided to the user through a presentation method. The response is displayed on the user's device and used as necessary information.

[1104] Output: The response displayed on the user's terminal

[1105] (Application example 1)

[1106] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1107] In recent years, customer targeting has become more sophisticated in the advertising industry, creating a demand for efficient ad generation that utilizes a variety of data. However, traditional ad generation methods require manual copywriting and targeting, which is time-consuming and labor-intensive, and it is difficult to generate consistent ad copy. To solve this problem, a system is needed that utilizes corporate and market data and automatically generates accurate ad copy based on user prompts.

[1108] 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.

[1109] In this invention, the server includes a data storage means, a preprocessing means, a model training means, and an advertising copy generation means, which preprocesses company data and market data and enables the automatic generation of consistent and highly accurate advertising copy.

[1110] "Data storage means" means means for holding basic data, company-specific data, academic knowledge data, multilingual data, legal data, general knowledge data and market data.

[1111] The "preprocessing means" is a means for reading data from the data storage means, processing missing values, deleting unnecessary columns, and filtering the data.

[1112] A "model training means" is a means for training a natural language processing model using preprocessed data and constructing a generative AI model.

[1113] "Generating means" means using a trained generative AI model to generate an optimal response based on prompts from a user.

[1114] The "advertising copy generation means" is a means for automatically generating consistent and highly accurate advertising copy based on company data and market data in response to user prompts.

[1115] "Company Data" means proprietary data related to a specific company that is used for targeting and personalization in advertising generation.

[1116] "Market Data" means data related to a particular market or industry and used for trend analysis and targeting.

[1117] This invention is a system for automatically generating advertising copy, comprising a data storage means, a preprocessing means, a model training means, a generation means, and an advertising copy generation means, which utilizes company data and market data to generate accurate advertising copy based on specific prompts from a user.

[1118] Data Storage Means

[1119] The server includes data storage means for holding basic data, company-specific data, academic knowledge data, multilingual data, legal data, general knowledge data, and market data. This data storage means includes hard disk drives and database systems required for data capture and management.

[1120] Pretreatment means

[1121] The server has a preprocessing means for reading data from the data storage means, processing missing values, deleting unnecessary items, filtering data, etc. For example, the server uses a data processing tool such as the Pandas library to efficiently preprocess the data.

[1122] Model training methods

[1123] Using the preprocessed data, a model training method builds a generative AI model, which uses a machine learning framework such as Python's Transformers library or TensorFlow for training.

[1124] generation means

[1125] The generator uses a trained generative AI model to generate an optimal response based on the user's prompts. Specifically, it uses the OpenAI API to invoke a natural language processing model and generate text based on the user's request.

[1126] Ad copy generation method

[1127] The advertising copy generation means automatically generates targeted advertising copy based on a prompt text provided by a user, utilizing company data and market data.

[1128] For example, if a user inputs a prompt such as "Please write an ad copy for a new dress from a fashion brand aimed at women in their 20s" on their smartphone or computer, the server tokenizes the prompt and feeds it to the trained model. The generated ad copy is then displayed.

[1129] For example, consider the following prompt:

[1130] "Please write an ad copy for a new dress from a fashion brand aimed at women in their 20s."

[1131] This allows users to quickly generate consistent and accurate ad copy and effectively develop advertising campaigns. In this way, the system of the present invention significantly improves the efficiency and accuracy of ad generation.

[1132] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1133] Step 1:

[1134] A user accesses the system and enters the API key and data path.

[1135] Input: API key, basic data path, corporate data path

[1136] Output: An initialized CustomAIModule instance.

[1137] How it works: A user logs into the system from a smartphone or PC and enters their OpenAI API key and data path. The server receives this information and initializes the instance.

[1138] Step 2:

[1139] The server reads the data from the data storage means.

[1140] Input: Basic Data Path, Corporate Data Path

[1141] Output: Data frame

[1142] What it does: The server uses the Pandas library to read the CSV file from the specified path and convert the base and company data into a dataframe.

[1143] Step 3:

[1144] The server executes the preprocessing means to process the data.

[1145] Input: DataFrame

[1146] Output: Preprocessed dataset

[1147] What it does: The server preprocesses the data frame, handling missing values, removing unnecessary items, and filtering the data, resulting in a clean, unified dataset.

[1148] Step 4:

[1149] The server uses the preprocessed data to perform model training.

[1150] Input: Preprocessed dataset

[1151] Output: A trained generative AI model

[1152] How it works: The server uses the Transformers library and TensorFlow to train a generative AI model using a preprocessed dataset. After training is complete, a highly accurate generative AI model is created.

[1153] Step 5:

[1154] The user enters a prompt statement.

[1155] Input: prompt statement

[1156] Output: Tokenized prompt

[1157] How it works: The user inputs a specific prompt, such as "Please write an ad copy for a new dress from a fashion brand aimed at women in their 20s," from their smartphone or computer. The server receives this prompt and performs a tokenization process.

[1158] Step 6:

[1159] The server generates the advertisement copy using the generation means.

[1160] Input: Tokenized prompt

[1161] Output: Generated ad text

[1162] How it works: The server feeds the tokenized prompts to a trained generative AI model to generate optimal ad copy, which is then displayed to the user.

[1163] Step 7:

[1164] The user reviews and uses the generated ad copy.

[1165] Input: Generated ad text

[1166] Output: Available ad text

[1167] How it works: Users review the ad copy provided by the system, modify or edit it as needed, and use it as the final ad copy, allowing for efficient and consistent ad campaigns.

[1168] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1169] The system of the present invention includes a data storage means, a pre-processing means, a model training means, a generation means, and an emotion engine.

[1170] First, the user configures the system by entering the OpenAI API key, the base data path, and the company's proprietary data path, and then initializes an instance of the CustomAIModule class. This gets the system ready to run.

[1171] The server accesses the data storage means that holds the basic data and the company's proprietary data, and reads the CSV files from each data path. The pre-processing means performs data cleansing and filtering on the data to generate an integrated data set.

[1172] The server then executes a model training process using the preprocessed data, which includes the basic data, the company-specific data, and the emotion information data, to create a highly accurate text generation model with emotion recognition capabilities.

[1173] After the model is trained, users can input specific prompts into the system, such as "Please give me some advice on how to respond to a customer complaint." The server tokenizes the prompt and provides it to the emotion engine, which analyzes the prompt content and classifies the user's emotion.

[1174] The server then feeds the emotion information classified by the emotion engine to a trained text generation model, and uses a generation means to generate an optimal response text, which includes an appropriate response according to the emotion.

[1175] As a concrete example, consider a customer support scenario. The user (customer support representative) sets the AI's API key, basic data path, and company-specific data (e.g., past customer interaction history). The server reads and preprocesses the basic data, company-specific data, and past customer interaction data to create a training dataset.

[1176] The server trains the model using the training data and displays a message indicating completion of training. When a customer support representative inputs "What should I do if a customer is angry about a product?" into the generative AI, the server analyzes the input prompt with an emotion engine and classifies the emotion. The server generates a response text according to the emotion and provides it to the representative.

[1177] In this way, the system of the present invention is able to recognize the user's emotions and quickly provide highly accurate and reliable information in response to the emotions.

[1178] The processing flow will be explained below.

[1179] Step 1:

[1180] The user sets the OpenAI API key, basic data path, and the company's own data path, and initializes an instance of the CustomAIModule class.

[1181] Step 2:

[1182] The server accesses a data storage means that holds basic data and company-specific data, and reads CSV files from each data path.

[1183] Step 3:

[1184] The server uses preprocessing tools to perform data cleansing on the loaded data, including handling missing values ​​and removing unnecessary columns.

[1185] Step 4:

[1186] The server combines the base data and the company's proprietary data into one integrated data set, so that all data is in a unified format.

[1187] Step 5:

[1188] The server executes a model training means to train a text generation model using the preprocessed data, which involves extracting and tokenizing text data from the dataset.

[1189] Step 6:

[1190] The server adds emotion information to the training data and creates a new training set that includes information corresponding to emotions.

[1191] Step 7:

[1192] The server begins training the text generation model using the tokenized text data and sentiment information. The training process is configured to span multiple epochs, during which the model learns the association between text features and sentiment.

[1193] Step 8:

[1194] The server confirms that training is complete and displays a message to the user saying "Model training is complete."

[1195] Step 9:

[1196] The user enters a specific prompt into the system (e.g., "What should I do if a customer is upset about my product?").

[1197] Step 10:

[1198] The server tokenizes the input prompt and feeds it to the emotion engine, which analyzes the user's input and classifies the emotion.

[1199] Step 11:

[1200] The server supplies the emotion information classified by the emotion engine to a trained text generation model, and generates an optimal response text using a generation means.

[1201] Step 12:

[1202] The server decodes the generated response text and presents it to the user.

[1203] Step 13:

[1204] The user checks the response text provided and uses it in their work as needed.

[1205] Example 2

[1206] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1207] Currently, there is a lack of systems that can accurately recognize customer emotions and quickly provide appropriate responses. This results in a problem of reduced efficiency of human resources and time in customer service operations in various fields, where efficient and effective support is required. In light of this situation, there is a need to develop a system that can achieve highly accurate emotion recognition and response generation.

[1208] 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.

[1209] In this invention, the server includes a user setting means, a data storage means, a preprocessing means, a model training means, a sentiment analysis means, and a text generation means, which enable the server to recognize the user's sentiment and quickly provide highly accurate and reliable information in response to the sentiment.

[1210] The "user setting means" is a function that allows the user to input setting information required to start the system operation.

[1211] "Data storage means" refers to a storage device or system for holding basic data, company-specific data, and sentiment information data.

[1212] The "preprocessing means" is a function that performs missing value complementation, correction of inconsistent data, and filtering of the acquired data to generate a clean dataset.

[1213] The "model training means" is a function that uses preprocessed data to train a machine learning model and create a highly accurate text generation model.

[1214] The "emotion analysis means" is a function that analyzes the user's emotions from the input text data and assigns appropriate emotion labels.

[1215] The "text generation means" is a function that generates optimal response text based on the emotion data obtained by the emotion analysis means.

[1216] The system of the present invention realizes highly accurate text generation by allowing the user to configure the system and input specific prompt sentences. Specifically, the system includes the following means: user configuration means, data storage means, preprocessing means, model training means, sentiment analysis means, and text generation means.

[1217] First, the user enters the necessary configuration information to set up the system, including the API key, the base data path, and the company's own data path. This initializes an instance of the CustomAIModule class, and the system is ready to run.

[1218] The server accesses the data storage means and reads the basic data, company-specific data, and emotion information data in CSV format. The data storage means refers to a storage device for storing these data.

[1219] Next, the server preprocesses the data using a preprocessing method, which performs missing value imputation, inconsistent data correction, duplicate data removal, and data filtering to generate a clean dataset. Specifically, the server performs data cleansing and filtering on the loaded data.

[1220] The server then uses a model training means to train a machine learning model based on the preprocessed data. The training data includes basic data, company-specific data, and sentiment information data. The training process is repeated to build a highly accurate text generation model. When training is complete, the server displays a message indicating that training is complete.

[1221] The user can input a specific prompt to the system. For example, "Please give me some advice on how to respond to a customer complaint." The server tokenizes this input and provides it to the sentiment analyzer. The sentiment analyzer analyzes the prompt content and classifies the user's sentiment.

[1222] Next, the server supplies the emotion information obtained by the emotion analysis means to the trained text generation model, and uses the text generation means to generate an optimal response text, which includes an appropriate response according to the emotion.

[1223] A concrete example of this would be a customer support scenario. The user (customer support representative) sets an API key, a basic data path, and the company's proprietary data (e.g., past customer interaction history). The server loads and preprocesses the basic data, the company's proprietary data, and past customer interaction data to create a training dataset. The server then trains the model using the training data, and a message indicating training completion is displayed. When the customer support representative inputs the following into the generative AI: "What should I do if a customer is upset about a product?", the server analyzes this input using the emotion engine and assigns an appropriate emotion label. Finally, the server generates the optimal response text based on the emotion and provides it to the representative.

[1224] In this way, the system of the present invention is able to recognize the user's emotions and quickly provide highly accurate and reliable information in response to the emotions.

[1225] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1226] Step 1:

[1227] The user configures the system.

[1228] Input: OpenAI API key, basic data path, company's proprietary data path

[1229] Operation: The user enters this information on the screen and sends it to the system. An instance of the CustomAIModule class is initialized and the system is ready to operate.

[1230] Output: System completed initialization

[1231] Step 2:

[1232] The server accesses the data storage means and reads the data.

[1233] Input: Basic data path, company's proprietary data path

[1234] Operation: Accesses the data storage means and reads the CSV files from each data path.

[1235] Output: Raw dataset of loaded basic and company data

[1236] Step 3:

[1237] The server pre-processes the data using the pre-processing means.

[1238] Input: Raw dataset

[1239] What it does: It performs data cleansing and filtering on the data it loads, including imputing missing values, correcting inconsistent data, removing duplicates, and filtering data.

[1240] Output: A clean dataset

[1241] Step 4:

[1242] The server trains the model using the model training means.

[1243] Input: Clean dataset, basic data, company data, sentiment data

[1244] How it works: A machine learning algorithm is used to train a model. The training process is repeated to build a highly accurate text generation model.

[1245] Output: A fully trained, highly accurate text generation model

[1246] Step 5:

[1247] The user enters a specific prompt.

[1248] Input: Prompt text (e.g., "Please advise how to handle a customer complaint.")

[1249] Action: Sends the entered prompt to the system.

[1250] Output: The prompt text entered into the system

[1251] Step 6:

[1252] The server parses the prompt.

[1253] Input: prompt statement

[1254] How it works: The prompt sentence is tokenized and fed to the sentiment analyzer, which analyzes the prompt content and classifies the user's sentiment.

[1255] Output: Classified emotion data

[1256] Step 7:

[1257] The server generates a response text based on the emotion information.

[1258] Input: Classified emotion data, trained text generation model

[1259] Operation: Generate optimal response text based on emotional information obtained by the emotion analysis means. Generate highly accurate text using the generation means.

[1260] Output: Optimal response text depending on the sentiment

[1261] Step 8:

[1262] The server provides the response text to the user.

[1263] Input: Response text

[1264] Behavior: The server generates a response text and displays it on the user's screen.

[1265] Output: The response text provided to the user

[1266] (Application example 2)

[1267] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1268] Conventional customer support systems have difficulty in recognizing and responding to users' emotions appropriately, making it difficult to increase user satisfaction, especially on online shopping sites. Returning a uniform response without considering emotions can further increase user dissatisfaction.

[1269] 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.

[1270] In this invention, the server includes a data storage means, a preprocessing means, a model training means, a user response generation means, and an emotion recognition means, which enable the server to analyze prompts input by a user, appropriately recognize the emotion, and generate a highly accurate response accordingly.

[1271] The "data storage means" is a storage device for storing data entered by users, academic knowledge data, multilingual data, legal data, general knowledge data, and company-specific data.

[1272] The "preprocessing means" is a device or software that performs cleansing such as processing missing values, deleting unnecessary columns, and filtering data on the data acquired from the data storage means, and integrates the data.

[1273] A "model training means" is a device or software for training a text generation model using preprocessed data.

[1274] A "generator" is a device or software that uses a trained model to generate response text based on a user's input.

[1275] An "emotion recognition means" is a device or software that analyzes the prompt entered by the user and classifies the emotion from its content.

[1276] The "user response generation means" is a device or software that generates an optimal response text based on the emotion information classified by the emotion recognition means.

[1277] A "prompt" is a question or request that a user enters into a system.

[1278] A "training dataset" is a collection of data that has been preprocessed for training a model.

[1279] "Emotion information" is information that indicates the type and intensity of the emotion contained in the user's prompt, as analyzed by the emotion recognition means.

[1280] A "text generation model" is a model that learns from a training dataset to generate appropriate text responses based on user prompts.

[1281] The system of the present invention includes a data storage means, a preprocessing means, a model training means, a generation means, an emotion recognition means, and a user response generation means. This system is specifically implemented as follows.

[1282] The system of the present invention first stores basic data entered by users, company-specific data, and emotion information data in a data storage means. These data are often saved as CSV files.

[1283] Next, the preprocessing means reads the data from the data storage means, handles missing values, removes unnecessary columns, and filters the data to generate a unified dataset, which provides consistent data for subsequent training processes.

[1284] The model training method uses preprocessed data to train a generative AI model. This training includes basic data, company-specific data, and sentiment data. The program is typically implemented using an existing language model API, such as the OpenAI API.

[1285] Once training is complete, the user response generator operates and accepts prompts from the user. For example, if the user inputs "A customer is angry about the product. What should I do?", the system analyzes the prompt with the emotion recognition unit and classifies the user's emotion.

[1286] The emotion information recognized by the emotion recognition means is supplied to the generation means, which generates an appropriate response text. The generated response is in line with the user's emotion and is provided with high accuracy and speed.

[1287] Specifically, when a user uses a customer support application on their smartphone and inputs, "My order hasn't arrived yet. What's going on?", the system first analyzes the input using emotion recognition and recognizes the emotion of "dissatisfaction (anger)." It then uses a generative AI model to generate an appropriate response text, providing the user with a support message such as, "We apologize for the delay in your order. We will check the current situation and get back to you as soon as possible."

[1288] Therefore, the system of the present invention can solve the problems that conventional systems have had and provide highly accurate customer support that responds to the user's emotions.

[1289] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1290] Step 1:

[1291] A user inputs a prompt sentence through a smartphone application, for example, "A customer is angry about the product. What should I do?" This input is sent to the server in text format.

[1292] Step 2:

[1293] The server reads the necessary data from the data storage means, which includes basic data, company-specific data, and sentiment data. The data is stored as CSV files, and the server parses these files.

[1294] Input: CSV file stored in your data storage means

[1295] Output: Raw data (basic data, company-specific data, sentiment data)

[1296] Step 3:

[1297] Preprocessing procedures are used on the raw data to handle missing values, remove unnecessary columns, and filter the data, creating a consistent, integrated dataset.

[1298] Input: Raw data read

[1299] Output: A consolidated dataset

[1300] Step 4:

[1301] The server uses a model training method to train a generative AI model using the pre-processed dataset, which is often done using an external language model API such as the OpenAI API.

[1302] Input: Preprocessed dataset

[1303] Output: A trained generative AI model

[1304] Step 5:

[1305] The server receives the prompt sentence entered by the user and analyzes it with the emotion recognition means, which analyzes the text of the prompt sentence and classifies the user's emotion.

[1306] Input: User prompt text

[1307] Output: Classified emotion information

[1308] Step 6:

[1309] The classified emotion information is supplied to the server's generation means, which generates a response text that matches the emotion. The generation means uses a trained generative AI model to generate the optimal response sentence.

[1310] Input: Classified emotion information, user prompt

[1311] Output: The generated response text

[1312] Step 7:

[1313] The response text generated by the response generating means is sent to the user terminal, and the user can view the generated response through a smartphone application.

[1314] Input: Generated response text

[1315] Output: The response message that is displayed to the user

[1316] The above steps enable highly accurate customer support that takes into account the user's emotions.

[1317] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1318] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1319] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1320] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1321] FIG. 9 illustrates an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and behaviors arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1322] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1323] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1324] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1325] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1326] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1327] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1328] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1329] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1330] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[1331] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1332] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1333] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1334] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1335] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1336] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1337] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1338] The following is further disclosed regarding the above embodiment.

[1339] (Claim 1)

[1340] data storage means;

[1341] A pre-processing means;

[1342] A model training means;

[1343] generating means;

[1344] A system including:

[1345] (Claim 2)

[1346] 2. The system according to claim 1, wherein the data storage means holds academic knowledge data, multilingual data, legal data, general knowledge data, and company data.

[1347] (Claim 3)

[1348] 2. The system of claim 1, wherein the preprocessing means reads the data from the data storage means, processes missing values, removes unnecessary columns, and filters the data.

[1349] (Claim 4)

[1350] 2. The system of claim 1, wherein the model training means trains the text generation model using preprocessed academic knowledge data, multilingual data, legal data, general knowledge data, and company-specific data.

[1351] (Claim 5)

[1352] 2. The system of claim 1, wherein the generating means generates the response using a text generation model trained based on user input.

[1353] "Example 1"

[1354] (Claim 1)

[1355] data storage means;

[1356] A pre-processing means;

[1357] A model training means;

[1358] generating means;

[1359] A setting input means for a user;

[1360] a prompt input means;

[1361] a response generating means;

[1362] A presentation means;

[1363] A system including:

[1364] (Claim 2)

[1365] 2. The system according to claim 1, wherein the data storage means holds academic knowledge data, multilingual data, legal data, general knowledge data, and company data.

[1366] (Claim 3)

[1367] 2. The system of claim 1, wherein the preprocessing means reads data from the data storage means, processes missing values, removes unnecessary columns, and filters the data to generate an integrated dataset.

[1368] (Claim 4)

[1369] 2. The system according to claim 1, wherein the user inputs settings by inputting an API key, a basic data path, and a company-specific data path, and initialization is performed based on the input information.

[1370] (Claim 5)

[1371] 2. The system of claim 1, wherein the prompt input means inputs specific questions or requests from the user, and the response generation means generates an optimal response to the input prompt using a trained model.

[1372] (Claim 6)

[1373] 6. The system according to claim 5, wherein the presenting means provides the generated response to the user.

[1374] "Application Example 1"

[1375] (Claim 1)

[1376] data storage means;

[1377] A pre-processing means;

[1378] A model training means;

[1379] generating means;

[1380] An advertisement copy generation means;

[1381] A system including:

[1382] (Claim 2)

[1383] 2. The system of claim 1, wherein the data storage means holds academic knowledge data, multilingual data, legal data, general knowledge data, company data, and market data.

[1384] (Claim 3)

[1385] 2. The system of claim 1, wherein the preprocessing means reads the data from the data storage means, processes missing values, removes unnecessary columns, and filters the data.

[1386] "Example 2: Combining Emotion Engines"

[1387] (Claim 1)

[1388] a user setting means;

[1389] data storage means;

[1390] A pre-processing means;

[1391] A model training means;

[1392] A sentiment analysis means;

[1393] a text generation means;

[1394] A system including:

[1395] (Claim 2)

[1396] 2. The system of claim 1, wherein the data storage means holds basic data, company-specific data, and sentiment information data.

[1397] (Claim 3)

[1398] 2. The system of claim 1, wherein the preprocessing means reads the data from the data storage means and performs missing value imputation, inconsistent data correction, and data filtering.

[1399] "Application example 2 when combining emotion engines"

[1400] (Claim 1)

[1401] data storage means;

[1402] A pre-processing means;

[1403] A model training means;

[1404] generating means;

[1405] An emotion recognition means;

[1406] A user response generating means;

[1407] A system including:

[1408] (Claim 2)

[1409] 2. The system according to claim 1, wherein the data storage means stores academic knowledge data, multilingual data, legal data, general knowledge data, and company-specific data, and the user response generation means generates a response text based on emotion information classified by the emotion recognition means.

[1410] (Claim 3)

[1411] 2. The system of claim 1, wherein the preprocessing means reads data from the data storage means, processes missing values, deletes unnecessary columns, and filters the data, and the emotion recognition means analyzes prompts input by a user and classifies emotions. [Explanation of symbols]

[1412] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. data storage means; A pre-processing means; A model training means; generating means; A system including:

2. 2. The system according to claim 1, wherein said data storage means holds academic knowledge data, multilingual data, legal data, general knowledge data, and company data.

3. 2. The system of claim 1, wherein the preprocessing means reads the data from the data storage means, processes missing values, removes unnecessary columns, and filters the data.

4. The system according to claim 1 , wherein the model training means performs training of the text generation model using preprocessed academic knowledge data, multilingual data, legal data, general knowledge data, and company-specific data.

5. 2. The system of claim 1, wherein the generating means generates responses using a text generation model trained based on user input.

Images