system

Abstract

Provide a system. A means for receiving parliamentary recording data; means for analyzing the received recorded data and converting the audio data into text data; means for formatting the converted text data into a predetermined format; a means for storing the formatted text data; A means for generating draft responses using a generative AI model based on previously stored text data; A means for outputting the generated answer plan; A system including:

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] Local government assembly work is becoming more complex every year, and a decrease in staff numbers is increasing the workload. Preparing assembly responses in particular requires a significant amount of time and effort, hindering efficient work execution. Furthermore, due to a lack of proper use of past assembly minutes, there is also the inefficiency of repeatedly creating responses to similar questions from scratch. Current generative AI models are unable to accurately and effectively understand existing minutes, making it difficult to generate the desired responses. [Means for solving the problem]

[0005] To solve the above-mentioned problems, the present invention provides a system including means for receiving parliamentary recording data, analyzing the recording data, and converting the audio data into text data, means for formatting the converted text data into a predetermined format, means for saving the formatted text data, means for automatically generating draft responses using a generative AI model based on previously saved text data, and means for outputting the generated draft responses. The system also includes means for accepting search queries in natural language, searching past minutes data based on the queries, and outputting the search results, making it possible to quickly and efficiently obtain the necessary information and create appropriate draft responses. The system also includes parliamentary recording and video recording equipment for handling parliamentary recording and video data.

[0006] "Parliamentary recording data" refers to data that records in audio format statements and discussions that took place in parliament.

[0007] "Means for receiving" refers to the function or device for collecting parliamentary recording data and incorporating it into the system.

[0008] The "means for analyzing and converting voice data into text data" refers to a function or device for converting voice data into text data in sentence format using voice recognition technology.

[0009] "Formatting means" refers to a function or device that formats the converted text data into a specific format (e.g., JSON format) that is easy for the generative AI model to use.

[0010] "Storage means" refers to a function or device for storing formatted text data in a database, cloud storage, etc.

[0011] A "generative AI model" is an artificial intelligence algorithm that learns from past data and generates new data, and in this case is specifically used to generate draft parliamentary responses.

[0012] "Means for generating a draft response" refers to a function or device for using a generative AI model to create a new draft response based on past minutes data.

[0013] "Output means" refers to a function or device that displays, prints, or outputs the generated answer plan in any other format to provide it to the user.

[0014] A "search query" is a question or request that a user enters in natural language to search for information in a database of past meeting minutes.

[0015] A "searching means" is a function or device for searching a database based on an input search query and extracting relevant information.

[0016] "Parliamentary recording equipment" means a physical device used to record audio and video of speeches and debates in parliament.

[0017] "Natural language" is the language humans use in everyday conversation, not a specific programming language or code. [Brief explanation of the drawings]

[0018] [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

[0019] 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.

[0020] First, the terms used in the following description will be explained.

[0021] 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).

[0022] 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.

[0023] 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.

[0024] 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.

[0025] 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."

[0026] [First embodiment]

[0027] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0028] 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.

[0029] 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).

[0030] 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.

[0031] 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.

[0032] 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.

[0033] 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.

[0034] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0035] 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.

[0036] 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.

[0037] 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.

[0038] 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."

[0039] The system of the present invention uses a generative AI model to create minutes of parliamentary sessions and automatically generate materials for the next session (such as draft answers) based on that data. Below, we will explain how this system is implemented in concrete terms.

[0040] Acquisition of parliamentary recording data

[0041] The user uploads a new session recording file to the server. For example, after a session ends, the user uploads the recording data (e.g., "session_2023_10.wav") via the web interface. The file is received by the server and stored in cloud storage.

[0042] Data analysis and minutes preparation

[0043] The server analyzes the uploaded recording file. A speech recognition engine is used for this analysis, converting the audio data into text data. Speech recognition technology such as DeepSpeech or Google (registered trademark) Speech-to-Text API is applied. The converted text data is split into segments for each speaker and appropriately formatted using speaker identification technology. It is then converted into a format (e.g., JSON) that is easy for the generative AI model to understand. Below is an example:

[0044] json

[0045] {

[0046] "session_id": "2023_10",

[0047] "timestamp": "2023-10-15",

[0048] "content": [

[0049] {

[0050] "speaker": "Rep. A",

[0051] "statement": "I'd like to ask about this year's budget."

[0052] },

[0053] {

[0054] "speaker": "Mayor B",

[0055] "statement": "The reason for the budget cuts is..."

[0056] }

[0057] ]

[0058] }

[0059] Storage of past meeting minutes data

[0060] The server saves the formatted minutes data in cloud storage. When saving, metadata (e.g., session ID, date) is also saved to ensure association with past minutes data.

[0061] Generate next parliamentary response

[0062] The user inputs the topic and questions for the next assembly meeting on the terminal. For example, the user might input, "I plan to ask a question about the budget at the next assembly meeting." The server then searches past minutes data based on this input and extracts relevant information.

[0063] The server generates new draft answers using a generative AI model (e.g., GPT-4 (registered trademark)) based on the extracted information. The generated draft answers are displayed on the device for the user to review and edit. For example, the following draft answers may be generated:

[0064] json

[0065] {

[0066] "suggestions": [

[0067] {

[0068] "topic": "budget",

[0069] "answer": "The budget cuts made last year were made to improve efficiency and were implemented in a way that minimized the impact on the educational environment."

[0070] }

[0071] ]

[0072] }

[0073] Natural language search function

[0074] A user can search the database of past meeting minutes using a natural language search query. For example, they can type "Show me last year's education budget discussion." The server analyzes this natural language query and searches for relevant meeting minutes data. As a result, the following search results are displayed on the user's device:

[0075] text

[0076] Meeting on October 15, 2022:

[0077] Councilman A: "Please explain last year's education budget cuts."

[0078] Mayor B: "The reason behind the budget cuts is..."

[0079] Parliamentary recording and video recording equipment

[0080] Users can also use audio and video recording equipment specifically designed for parliamentary use. This equipment is designed to record high-quality audio and video recordings of parliamentary speeches and debates. The audio and video recordings are then uploaded to a server for later analysis and storage.

[0081] This system allows local government employees to carry out their work efficiently and significantly reduces the time it takes to prepare assembly response proposals. In addition, by utilizing past meeting minutes data, it is possible to prepare consistent response proposals.

[0082] The processing flow will be explained below.

[0083] Step 1:

[0084] The user uploads the parliamentary recording (e.g., "session_2023_10.wav") to the server using the interface. The user selects the recording using a web browser and clicks the "Upload" button.

[0085] Step 2:

[0086] The server receives the uploaded recording data. After receiving it, the file is stored in cloud storage, where the file's metadata (e.g., session ID, upload date and time) is also recorded.

[0087] Step 3:

[0088] The server analyzes the received recording data, activates a speech recognition engine, and converts the audio data into text data, which is then stored in a temporary database.

[0089] Step 4:

[0090] The server extracts the text data obtained from the speech recognition engine and uses speaker identification technology to segment it into segments for each speaker, identifying speakers based on, for example, voice patterns or time of day.

[0091] Step 5:

[0092] The server formats the text data of the identified speaker into a format that is easy for the AI ​​model to understand (e.g., JSON format). The formatted data is then stored in cloud storage. The following is an example of the data:

[0093] json

[0094] {

[0095] "session_id": "2023_10",

[0096] "timestamp": "2023-10-15",

[0097] "content": [

[0098] {

[0099] "speaker": "Rep. A",

[0100] "statement": "I'd like to ask about this year's budget."

[0101] },

[0102] {

[0103] "speaker": "Mayor B",

[0104] "statement": "The reason for the budget cuts is..."

[0105] }

[0106] ]

[0107] }

[0108] Step 6:

[0109] The user enters the topic and question for the next session of the Diet on the device. For example, they enter, "I plan to ask a question about the budget at the next session of the Diet." The input is done through a web interface.

[0110] Step 7:

[0111] The server searches for past minutes data based on the input topic, and efficiently extracts related minutes data using metadata.

[0112] Step 8:

[0113] The server uses a generative AI model (e.g., GPT-4) to generate a new draft response based on the extracted minutes data. Below is an example of a generated draft response:

[0114] json

[0115] {

[0116] "suggestions": [

[0117] {

[0118] "topic": "budget",

[0119] "answer": "The budget cuts made last year were made to improve efficiency and were implemented in a way that minimized the impact on the educational environment."

[0120] }

[0121] ]

[0122] }

[0123] Step 9:

[0124] The generated draft answer is displayed on the device, where the user can review it and edit it if necessary. The final draft answer can also be saved in a different format or printed.

[0125] Step 10:

[0126] A user enters a natural language search query, for example, "Show me last year's education budget discussions."

[0127] Step 11:

[0128] The server analyzes the entered search query and searches the database for relevant meeting minutes data, using natural language processing techniques.

[0129] Step 12:

[0130] The search results are displayed on the device. For example, the following search results are displayed on the user's screen:

[0131] text

[0132] Meeting on October 15, 2022:

[0133] Councilman A: "Please explain last year's education budget cuts."

[0134] Mayor B: "The reason behind the budget cuts is..."

[0135] Example 1

[0136] 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."

[0137] Conventional methods for managing parliamentary recording data require a great deal of time and effort for manual minutes creation and data search, resulting in low productivity. It is also difficult to quickly propose consistent content for the next parliamentary session. Therefore, there is a need for a system that can automatically and efficiently create minutes and provide draft answers using generative AI models.

[0138] 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.

[0139] In this invention, the server includes means for receiving recorded data, means for analyzing the received recorded data and converting the audio data into text data, means for formatting the converted text data into a predetermined format, means for saving the formatted text data, means for generating a draft answer using a generative AI model based on previously saved text data, means for outputting the generated draft answer, and means for displaying the output draft answer on a terminal for user input. This makes it possible to automatically and efficiently create minutes and quickly propose consistent draft answers using the generative AI model.

[0140] "Recorded data" refers to data that records in audio format statements and discussions at meetings such as parliaments.

[0141] "Means for receiving" refers to the function or device that allows the server to import the recorded data uploaded by the user.

[0142] "Means for analyzing" refers to the technology or method for analyzing the received recorded data and converting the audio data into text data.

[0143] "Audio data" refers to audio signals stored as recorded data.

[0144] "Text data" is data of character strings that are obtained by analyzing voice data and converting it into linguistic information.

[0145] The "predetermined format" refers to the form or structure used to format the generated text data to ensure consistency and ease of understanding.

[0146] "Storage means" refers to a function or system for storing and managing formatted text data.

[0147] A "generative AI model" refers to a model that uses artificial intelligence to generate responses or suggestions based on specific input data.

[0148] A "draft answer" is a prepared answer that a member of parliament or other relevant person can use to appropriately respond to a question asked in a parliament or other such setting.

[0149] "Means of output" refers to the functions and systems for presenting the generated text data and draft answers to the user.

[0150] A "user input terminal" is a device used by a user to enter data and check results.

[0151] A "natural language search query" refers to a question that a user enters in natural language to search past data.

[0152] "Query-based search means" refers to techniques or methods for retrieving relevant past data from an input natural language search query.

[0153] "Means for outputting search results" refers to the functions and systems for displaying search results to the user.

[0154] "Recording device" refers to a device or equipment for acquiring audio and video data.

[0155] This invention is a system that uses generative AI models to efficiently create parliamentary minutes and automatically generate materials for the next parliamentary session. The system starts with the user uploading audio data, then analyzes the audio data, converts it to text data, stores the data, searches it, and generates draft responses.

[0156] Acquisition of parliamentary recording data

[0157] A user uploads the recorded data to the server via a web interface. For example, after a session ends, the user selects the recorded data (e.g., "session_2023_10.wav") and clicks the upload button. The server receives the data as an HTTP POST request and saves it to cloud storage (e.g., Amazon S3). The server notifies the user when the file has been saved.

[0158] Data analysis and minutes preparation

[0159] The server passes the received recording data to a speech recognition engine, which converts the audio data into text data. This analysis is performed using, for example, the Google Speech-to-Text API. The audio data is converted into a byte stream and sent via the API. The returned text data is split into segments for each speaker using speaker identification technology and formatted into a JSON format that is easy for the generative AI model (e.g., GPT-4) to understand.

[0160] For example, the following JSON data is generated:

[0161] json

[0162] {

[0163] "session_id": "2023_10",

[0164] "timestamp": "2023-10-15",

[0165] "content": [

[0166] {

[0167] "speaker": "Rep. A",

[0168] "statement": "I'd like to ask about this year's budget."

[0169] },

[0170] {

[0171] "speaker": "Mayor B",

[0172] "statement": "The reason for the budget cuts is..."

[0173] }

[0174] ]

[0175] }

[0176] Storage of past meeting minutes data

[0177] The server saves the formatted JSON data in cloud storage. This allows it to be associated with past meeting minutes data. When saving, metadata (e.g., session ID, date) is also saved.

[0178] Generate next parliamentary response

[0179] The user uses an input form on their device to input the topic and questions for the next session. For example, they might input, "I plan to ask a question about the budget at the next session." The server receives this input and searches past minutes data to extract relevant information. It then uses a generative AI model (e.g., GPT-4) to generate new draft responses. An example of this prompt is, "Imagine a question about 'budget cuts' at the next session. Please generate a relevant draft response based on past minutes data." The generated draft response is displayed on the user's device.

[0180] Natural language search function

[0181] Users can search past meeting minutes data by entering natural language search queries into their devices. For example, they can enter "Show me last year's education budget discussions." The server analyzes this input using a natural language processing (NLP) module and searches for relevant data. Search results are displayed on the user's device.

[0182] You'll see specific search results like this:

[0183] text

[0184] Meeting on October 15, 2022:

[0185] Councilman A: "Please explain last year's education budget cuts."

[0186] Mayor B: "The reason behind the budget cuts is..."

[0187] Parliamentary recording and video recording equipment

[0188] Users operate dedicated audio and video recording equipment for council meetings to record speeches and discussions in high quality. The recorded data is later uploaded to a server for analysis and storage. This system enables local government officials to carry out their work efficiently and prepare consistent responses.

[0189] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0190] Step 1: Upload your recordings

[0191] The user uploads the recording (e.g., "session_2023_10.wav") to the server via a web interface. Input: The recording file of the parliament. Output: The uploaded recording file is saved in cloud storage. Specific behavior: The user opens a file selection dialog in a web browser, selects the recording data from the local file system, and presses the upload button. The server receives the HTTP POST request and saves the file in cloud storage.

[0192] Step 2: Analyzing the audio data

[0193] The server retrieves the recorded data from cloud storage and sends it to a speech recognition engine (e.g., Google Speech-to-Text API). Input: The saved recording file. Output: The speech data converted to text data. Specific operation: The server converts the speech data into a byte stream, sends it to the recognition engine, and receives the returned text data.

[0194] Step 3: Formatting the text data

[0195] The server analyzes the acquired text data and uses speaker identification technology to split and format it for each speaker. Input: Text data returned from the speech recognition engine. Output: Meeting minutes data in JSON format. Specific operation: The server analyzes the text data, splits it for each speaker, and generates formatted JSON data.

[0196] Step 4: Save your data

[0197] The server saves the formatted JSON data in cloud storage. Input: Formatted text data (JSON format). Output: Meeting minutes data saved in cloud storage. Specific operation: The server saves the JSON data in the specified directory, along with the metadata.

[0198] Step 5: Generate draft answers for the next session

[0199] The user inputs the topic and question for the next session of the Diet (e.g., "I plan to ask a question about the budget at the next session of the Diet") into the device. Input: The topic and question for the next session of the Diet. Output: Generated draft answers. Specific operation: The server searches past minutes data based on the input topic and question, generates new draft answers using a generative AI model (e.g., GPT-4), and displays them on the device.

[0200] Step 6: Natural Language Search

[0201] A user enters a natural language search query (e.g., "Show me last year's education budget discussions") into a device. Input: Natural language search query. Output: Relevant meeting minutes data. Specific behavior: The server uses a natural language processing (NLP) module to parse the search query, search for relevant data, and display the search results on the device.

[0202] Step 7: Operating the recording equipment

[0203] The user operates audio and video recording equipment specifically designed for the assembly to capture high-quality audio and video data. Input: Data captured by the audio and video recording equipment. Output: The audio and video data is uploaded to the server and stored in cloud storage. Specific operation: The user presses the start recording button on the audio and video recording equipment, which records the entire assembly meeting and then uploads the data to the server.

[0204] (Application example 1)

[0205] 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."

[0206] In logistics centers, creating minutes of meetings and conferences is a time-consuming and labor-intensive process, and labor shortages can lead to inefficiencies. It is also difficult to quickly retrieve information related to the next meeting or task by searching past meeting records. The present invention aims to solve these problems by providing a system that automatically converts meeting recording data into text and generates materials and plans for the next meeting based on a generative AI model.

[0207] 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.

[0208] In this invention, the server includes means for receiving recorded assembly data, means for analyzing the received recorded data and converting the audio data into text data, means for formatting the converted text data into a predetermined format, means for saving the formatted text data, means for generating draft responses using a generative AI model based on previously saved text data, means for inputting the next meeting theme and tasks, means for extracting related information and generating new meeting materials and draft plans, and means for outputting the generated draft responses and meeting materials. This automates the creation of meeting minutes within the logistics center, enabling the efficiency of preparations for the next meeting and tasks.

[0209] "Parliamentary recording data" refers to data files containing recordings of speeches and discussions held in parliaments or meetings.

[0210] "Means for receiving" refers to the function for obtaining recorded data from outside and importing it into the server.

[0211] "Means for analyzing and converting voice data into text data" refers to a function for converting recorded voice data into text information using voice recognition technology.

[0212] "Means for formatting into a specified format" refers to a function for arranging converted text data according to certain rules and formats.

[0213] "Means of storage" refers to the function of storing formatted text data in a database, cloud storage, etc. for future reference or search.

[0214] "Means for generating draft responses using generative AI models" refers to a function that uses AI technology based on past data to automatically create draft responses to be used in the next parliament or meeting.

[0215] "Means for inputting the next meeting topic and tasks" refers to an interface that allows the user to provide the server with information and questions about the next meeting.

[0216] "Means for extracting relevant information and generating new materials and plans" refers to the function of extracting necessary information from past data and using it to create new materials and plans.

[0217] "Means for outputting" refers to a function for providing the generated draft answers and meeting materials to the user.

[0218] "Means for accepting natural language search queries" refers to an interface that allows users to enter search terms in everyday language.

[0219] "Means for searching previously stored text data based on a search query" refers to a function for extracting data that matches the user's input from a database.

[0220] "Means for outputting search results" refers to a function for providing the searched content to the user.

[0221] "Means for additionally analyzing recorded data" refers to a function for further analyzing recorded data within a logistics facility and converting it into text data, etc.

[0222] "Function to record meeting recording data" refers to the function to collect the contents of meetings and assemblies in audio format.

[0223] "The function of recording recorded data" refers to the function of collecting the contents of meetings and assemblies in video format.

[0224] "Recording and video recording equipment" refers to devices and equipment used to record the contents of meetings and assemblies as audio or video.

[0225] MODE FOR CARRYING OUT THE INVENTION

[0226] This system automatically analyzes recorded meeting data at a logistics center, converts it into text data, and saves it. It can also automatically generate materials and plans for the next meeting using a generative AI model. It also has the ability to search past meeting records using natural language search queries and provide the results to users.

[0227] System Overview

[0228] The system consists of the following main components:

[0229] 1. Means of receiving parliamentary recordings

[0230] 2. Means of converting audio data into text data

[0231] 3. A means of formatting text data into a specified format

[0232] 4. Means of saving text data

[0233] 5. How to generate draft responses using generative AI models

[0234] 6. A way to enter the next meeting topic or task

[0235] 7. A means to extract relevant information and generate new meeting materials and plans

[0236] 8. Means for outputting the generated response plan and meeting materials

[0237] 9. A way to accept natural language search queries

[0238] 10. A way to search past data based on a search query

[0239] 11. How to output search results

[0240] 12. Additional means of analyzing recording data

[0241] 13. Function to record conference recording data

[0242] 14. Audio and video recording devices with the function of recording video data

[0243] System Operation

[0244] The server receives the parliamentary recording data, analyzes it, and converts the audio data into text data. This speech recognition uses technologies such as DeepSpeech and Google Speech-to-Text API. The converted text data is formatted and stored in cloud storage (e.g., AWS (registered trademark) S3).

[0245] Users can input the topic or task for the next meeting into their device. For example, they can input a prompt such as, "Please create a document to discuss the logistics situation at the next meeting." The server then searches past text data and extracts relevant information. Based on this extracted information, a generative AI model (e.g., GPT-4) is used to generate new meeting materials and draft plans. The generated materials and draft responses are then displayed on the user's device.

[0246] Additionally, users can search for past meeting records using natural language search queries, such as a prompt such as "Show me meeting records from last year regarding logistics status." The server analyzes the query, searches for relevant data, and provides the results to the user.

[0247] Hardware and software used

[0248] The system uses the following hardware and software:

[0249] Server: Amazon Web Services (AWS) EC2

[0250] Cloud storage: AWS S3

[0251] Speech recognition engine: DeepSpeech, Google Speech-to-Text API

[0252] Generative AI model: OpenAI® GPT-4

[0253] Natural language processing libraries: SpaCy, NLTK

[0254] Specific examples

[0255] For example, suppose a logistics center manager holds a weekly meeting and records it. The recording is then uploaded to a server, which analyzes the data and creates meeting minutes. For next week's meeting, if a user types into their device, "Please generate this week's plan based on last week's minutes," the server will search past data and automatically generate a new plan using a generative AI model. This can improve work efficiency.

[0256] Examples of prompts include "Please prepare materials to talk about the logistics situation at the next meeting" and "Show me the minutes of last year's meeting about the logistics situation."

[0257] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0258] Step 1:

[0259] A user acquires parliamentary recording data using audio and video recording equipment and uploads the data to the server. In this example, the recording file (e.g., "meeting_2023_10.wav") is sent to the server via a web interface. The input is an audio file, and the output is an audio file stored in cloud storage.

[0260] Step 2:

[0261] The server analyzes the received recording data and converts the audio data into text data. This process uses a speech recognition engine (DeepSpeech or Google Speech-to-Text API). The input is an audio file, and the output is the converted text data.

[0262] Step 3:

[0263] The server formats the converted text data into a specified format. Specifically, it uses speaker identification technology to split the text data for each speaker and converts it into JSON format. The input is raw text data, and the output is formatted text data.

[0264] Step 4:

[0265] The server saves the formatted text data in cloud storage (AWS S3), along with metadata (e.g., session ID, date). The input is the formatted text data, and the output is the data saved in cloud storage.

[0266] Step 5:

[0267] The user inputs the next meeting topic and task into the terminal. For example, the user might input, "Please prepare materials to talk about the logistics situation at the next meeting." The input is a prompt text, and the output is sending the prompt text to the server.

[0268] Step 6:

[0269] The server searches for information related to the prompt text entered by the user based on previously saved text data. The input is the prompt text and past text data, and the output is the searched related information.

[0270] Step 7:

[0271] Based on the related information retrieved by the server, a generative AI model (GPT-4) is used to generate meeting materials and plan proposals. The input is related information, and the output is new meeting materials and plan proposals.

[0272] Step 8:

[0273] The server outputs the generated meeting materials and plans to the terminals to provide them to the users. The input is the new meeting materials and plans, and the output is the materials and plans displayed on the user's terminal.

[0274] Step 9:

[0275] A user inputs a search query in natural language into a terminal. For example, "Show me the minutes of the meeting about the logistics situation last year." The input is a search query in natural language, and the output is a search query sent to the server.

[0276] Step 10:

[0277] The server searches past text data based on the search query. Here, it references a database of past meeting minutes and extracts relevant data. The input is the search query and past text data, and the output is the search result data.

[0278] Step 11:

[0279] The server outputs the search results to the terminal, allowing the user to view the results. The input is the search result data, and the output is the search result displayed on the user's terminal.

[0280] 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.

[0281] The system of the present invention uses a generative AI model to create minutes of parliamentary sessions and automatically generates the next parliamentary session materials (such as draft parliamentary responses) based on that data. Furthermore, by combining it with an emotion engine that recognizes the user's emotions, it is possible to generate more appropriate draft responses that reflect the user's intentions and emotions. Below, we will explain in detail how this system is implemented.

[0282] Acquisition of parliamentary recording data

[0283] The user uploads the parliamentary recording (e.g., "session_2023_10.wav") to the server using the interface. The user selects the recording via a web browser and clicks the "Upload" button. The file is received by the server and stored in cloud storage.

[0284] Data analysis and minutes preparation

[0285] The server analyzes the uploaded recording data, activates a speech recognition engine, and converts the audio data into text data. The speech recognition technology used is DeepSpeech or the Google Speech-to-Text API, for example. The converted text data is then stored in a temporary database.

[0286] Speaker Identification and Data Shaping

[0287] The server extracts the text data obtained from the speech recognition engine and divides it into segments for each speaker using speaker identification technology. For example, speakers can be identified based on voice patterns or time of day. The server then formats the text data of the identified speaker into a format (e.g., JSON) that is easy for the generative AI model to understand. The formatted data is then saved in cloud storage. The following is an example of the data:

[0288] json

[0289] {

[0290] "session_id": "2023_10",

[0291] "timestamp": "2023-10-15",

[0292] "content": [

[0293] {

[0294] "speaker": "Rep. A",

[0295] "statement": "I'd like to ask about this year's budget."

[0296] },

[0297] {

[0298] "speaker": "Mayor B",

[0299] "statement": "The reason for the budget cuts is..."

[0300] }

[0301] ]

[0302] }

[0303] Generating Answer Plans and Using an Emotion Engine

[0304] The user inputs the topic and questions for the next session of the Diet on the terminal. For example, the user might input, "I plan to ask a question about the budget at the next session of the Diet." The input is done via a web interface. The server searches past minutes data based on this input and extracts relevant information.

[0305] The server uses a generative AI model (e.g., GPT-4) to generate new response proposals based on the extracted information. Furthermore, an emotion engine recognizes the user's emotional data and adjusts the response proposals based on that data. For example, if the user is recognized as "nervous," the response proposals will include expressions that convey a more reassuring feeling.

[0306] json

[0307] {

[0308] "suggestions": [

[0309] {

[0310] "topic": "budget",

[0311] "answer": "The budget cuts made last year were made to improve efficiency and we have taken measures to minimize the impact on the educational environment. Please rest assured."

[0312] }

[0313] ]

[0314] }

[0315] Storing and using emotional data

[0316] The server stores the user's emotional data recognized by the emotion engine in cloud storage. The stored emotional data is used to generate future parliamentary response drafts. This results in more consistent response drafts and increases user confidence.

[0317] Natural language search function

[0318] A user can search the database of past meeting minutes using a natural language search query. For example, they can type "Show me last year's education budget discussion." The server analyzes this natural language query and searches for relevant meeting minutes data. As a result, the following search results are displayed on the user's device:

[0319] text

[0320] Meeting on October 15, 2022:

[0321] Councilman A: "Please explain last year's education budget cuts."

[0322] Mayor B: "The reason behind the budget cuts is..."

[0323] Parliamentary recording and video recording equipment

[0324] Users can record high-quality audio and video of parliamentary speeches and debates using dedicated parliamentary audio and video recording equipment, which will then be uploaded to a server for later analysis and storage.

[0325] This system allows local government employees to carry out their work efficiently and significantly reduces the time it takes to prepare assembly response drafts. Furthermore, by taking into consideration the user's feelings, it is possible to provide more reliable response drafts. In addition, by utilizing past minutes data, it is possible to prepare consistent response drafts.

[0326] The processing flow will be explained below.

[0327] Step 1:

[0328] The user uploads the parliamentary recording (e.g., "session_2023_10.wav") to the server using the interface. The user selects the recording via a web browser and clicks the "Upload" button.

[0329] Step 2:

[0330] The server receives the uploaded recording data. After receiving it, the file is stored in cloud storage, where the file's metadata (e.g., session ID, upload date and time) is also recorded.

[0331] Step 3:

[0332] The server analyzes the received recording, invokes a speech recognition engine (e.g., DeepSpeech or Google Speech-to-Text API), and converts the audio data into text data, which is then stored in a temporary database.

[0333] Step 4:

[0334] The server extracts the text data obtained from the speech recognition engine and uses speaker identification technology to segment it into segments for each speaker, identifying speakers based on, for example, voice patterns or time of day.

[0335] Step 5:

[0336] The server formats the text data of the identified speaker into a format that is easy for the AI ​​model to understand (e.g., JSON format). The formatted data is then stored in cloud storage. The following is an example of the data:

[0337] json

[0338] {

[0339] "session_id": "2023_10",

[0340] "timestamp": "2023-10-15",

[0341] "content": [

[0342] {

[0343] "speaker": "Rep. A",

[0344] "statement": "I'd like to ask about this year's budget."

[0345] },

[0346] {

[0347] "speaker": "Mayor B",

[0348] "statement": "The reason for the budget cuts is..."

[0349] }

[0350] ]

[0351] }

[0352] Step 6:

[0353] The user inputs the topic and questions for the next council meeting on the terminal. For example, the user inputs, "I plan to ask a question about the budget at the next council meeting." The input is done via a web interface.

[0354] Step 7:

[0355] The server searches for past minutes data based on the input topic, and efficiently extracts related minutes data using metadata.

[0356] Step 8:

[0357] The server uses a generative AI model (e.g., GPT-4) to generate a new draft response based on the extracted minutes data. Below is an example of a generated draft response:

[0358] json

[0359] {

[0360] "suggestions": [

[0361] {

[0362] "topic": "budget",

[0363] "answer": "The budget cuts made last year were made to improve efficiency and were implemented in a way that minimized the impact on the educational environment."

[0364] }

[0365] ]

[0366] }

[0367] Step 9:

[0368] The server uses the emotion engine to recognize the user's emotion, for example, the emotion engine recognizes that the user is "nervous."

[0369] Step 10:

[0370] The server reflects the emotional data it recognizes in the generated answer proposals. For example, if the server recognizes that the user is "nervous," the answer proposals include expressions that make the user feel more at ease.

[0371] Step 11:

[0372] The generated draft answer is displayed on the device, where the user can review it and edit it if necessary. The final draft answer can also be saved in a different format or printed.

[0373] Step 12:

[0374] A user enters a natural language search query, for example, "Show me last year's education budget discussions."

[0375] Step 13:

[0376] The server analyzes the entered search query and searches the database for relevant meeting minutes data, using natural language processing techniques.

[0377] Step 14:

[0378] The search results are displayed on the device. For example, the following search results are displayed on the user's screen:

[0379] text

[0380] Meeting on October 15, 2022:

[0381] Councilman A: "Please explain last year's education budget cuts."

[0382] Mayor B: "The reason behind the budget cuts is..."

[0383] Example 2

[0384] 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."

[0385] The task of creating minutes of assembly meetings and response materials for the next assembly meeting requires time and effort, placing a heavy burden on staff. Furthermore, creating response materials requires appropriately reflecting the content of past discussions while also taking into account the user's emotions and intentions, but there is no efficient way to do this. In conventional technologies, creating minutes and generating response materials is often done manually, resulting in reduced work efficiency and inconsistent quality. This invention is intended to solve these problems.

[0386] The specific processing by the specific processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for receiving parliamentary recording data, means for analyzing the received recording data and converting the audio data into text data, means for formatting the converted text data into a predetermined data format, means for generating a draft response using a generative AI model based on previously saved text data, means for reflecting the user's emotions in the generated draft response using an emotion engine, and means for outputting the generated draft response. This automates the creation of minutes and the generation of response materials, improving work efficiency and ensuring consistent quality. Furthermore, because response materials are generated taking the user's emotions into consideration, more appropriate and reliable materials can be created.

[0387] "Parliamentary recording data" refers to audio data of speeches and discussions in parliament.

[0388] A "generative AI model" refers to an artificial intelligence model that generates new outputs based on input data.

[0389] "Emotion engine" refers to a program or function that recognizes a user's emotions and analyzes the emotion data.

[0390] "Draft answers" refers to prepared answers to questions asked in parliament.

[0391] "Converted text data" refers to character data converted from voice data using voice recognition technology.

[0392] "Data format" refers to the format or standard for organizing and arranging data in a certain structure.

[0393] A "natural language search query" refers to a search request made to a database using natural language.

[0394] "Sound recording equipment" means devices for recording sound and video.

[0395] "Cloud storage" refers to an online storage service that stores and manages data via the Internet.

[0396] A "server" refers to a computer system that provides services and data to other computers over a network.

[0397] "Terminal" refers to a device used by a user for input and display.

[0398] "Text data" refers to data expressed as character information.

[0399] The system of the present invention utilizes a generative AI model and an emotion engine to create minutes of parliamentary meetings and automatically generate materials for responses to questions at the next parliamentary meeting. A detailed embodiment of this system will be described.

[0400] Overall system overview

[0401] This system analyzes recordings of parliamentary sessions, converts the audio data into text data, and uses that text data to automatically generate draft responses for the next session. It can also recognize the user's emotions and generate draft responses that reflect those emotions. This system consists of three main components: a server, a terminal, and a user.

[0402] Hardware and Software Configuration

[0403] The server receives the recorded data and analyzes the audio data. It is equipped with a high-performance processor and a large amount of memory. The speech recognition engine uses common speech recognition software such as DeepSpeech or Google Speech-to-Text API. The server also converts the analyzed text data into a specified data format and stores it in a temporary database.

[0404] The terminal provides an interface for users to input the topic and questions for the next session. This terminal is a standard PC or tablet and is connected to the server via a web browser. Users can use the terminal to upload recording data as well as check the generated response plans.

[0405] Users upload the recorded data of assembly meetings to the server using their devices. They input the topic and questions of the next assembly meeting, and check and use the proposed answers provided by the server. Users can also input search queries to search for past minutes data.

[0406] Processing flow and specific examples

[0407] The system begins by a user uploading a recording of a parliamentary session. The recording (e.g., "session_2023_10.wav") is then stored in cloud storage by the server. The server then activates a speech recognition engine, analyzes the recording, and converts the audio data into text data. This text data is then split by speaker using speaker identification technology and formatted into a format that is easy for the generative AI model to understand.

[0408] For example, a statement such as "I'd like to ask about this year's budget" is linked to "Council Member A," and a statement such as "The reason for the budget cuts is..." is linked to "Mayor B." This results in the data being formatted as text like this:

[0409] Session ID: 2023_10

[0410] Timestamp: 2023-10-15

[0411] Contents:

[0412] Councilman A: "I'd like to ask about this year's budget."

[0413] Mayor B: 'The reason for the budget cuts is...'

[0414] "

[0415] Next, the user inputs the next session's theme and question on the device. For example, they might input, "I plan to ask a question about the budget at the next session." The server extracts relevant information from past minutes data and uses this information to input prompts into the generative AI model to generate new draft responses. The emotion engine also recognizes the user's emotions, and if it recognizes that the user is "nervous," the generated draft responses will include expressions that convey a sense of security.

[0416] Examples of prompts are:

[0417] "Parliamentary Data:

[0418] Session ID: 2023_10

[0419] Timestamp: 2023-10-15

[0420] Contents:

[0421] Councilman A: "I'd like to ask about this year's budget."

[0422] Mayor B: 'The reason for the budget cuts is...'

[0423] Next Congress Theme and User Sentiments:

[0424] Theme: 'Budget'

[0425] Question: I plan to ask a question about the budget at the next Congress. I would like to know specifically why the budget cuts were made.

[0426] Emotion: 'Nervous'

[0427] Generated answer plan:

[0428] Topic: 'Budget'

[0429] Response: 'The budget cuts made last year were made to improve efficiency and we have taken measures to minimize the impact on the educational environment. Please rest assured.'

[0430] "

[0431] The server saves the generated answer plan in cloud storage and displays it on the user's device. The user's emotional data is also saved in cloud storage and used to generate answer plans from the next time onwards, making it possible to create consistent answer plans.

[0432] The above is a specific embodiment of the system. This system automates the creation of assembly minutes and the generation of response materials for the next assembly meeting, and can provide high-quality materials that take user emotions into consideration.

[0433] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0434] Step 1: Obtaining Congressional Recording Data

[0435] Input: The user uploads a recording file (e.g., "session_2023_10.wav") through the designated interface.

[0436] How it works: A user uses a web browser to select a recording and presses the "upload" button.

[0437] Output: The server saves this recording to cloud storage and records the file's metadata (e.g., file name, upload date and time).

[0438] Step 2: Analyze the data and create a transcript

[0439] Input: Recording data files stored in cloud storage.

[0440] How it works: The server launches a speech recognition engine (e.g., DeepSpeech or Google Speech-to-Text API) and reads the recording. The server converts the audio data into text. This conversion process converts what is being said into text format.

[0441] Output: The converted text data is saved in a temporary database.

[0442] Step 3: Speaker identification and data conditioning

[0443] Input: Text data stored in a temporary database.

[0444] How it works: The server uses speaker identification technology to identify each speaker, using characteristics such as voice patterns and time of day. It then organizes the text data for each identified speaker and formats it into a data format (e.g., JSON) that is easy for the generative AI model to understand.

[0445] Output: The formatted text data is saved to cloud storage.

[0446] Step 4: Enter the theme and questions for the next assembly meeting

[0447] Input: The user enters the topic or question for the next session through a web interface (e.g., "I have a question about the budget for the next session").

[0448] How it works: The user uses a terminal to enter the required information and send it to the server.

[0449] Output: The server receives this input data and prepares it for the next processing step.

[0450] Step 5: Generate a response and use the emotion engine

[0451] Input: User questions and past meeting minutes data.

[0452] How it works: The server extracts relevant information from past meeting minutes data and inputs prompt sentences into a generative AI model. A generative AI model (e.g., GPT-4) is used to generate new response proposals. The emotion engine then analyzes the user's emotion data and adjusts the response proposals based on the recognized emotion. For example, "If the user is nervous, add expressions that will make them feel at ease."

[0453] Output: A response plan reflecting the sentiment is generated and saved in cloud storage.

[0454] Step 6: Outputting the generated answer plan

[0455] Input: Response plan data stored on the server.

[0456] Operation: The server sends the generated answer plan to the user's terminal, where the user can view the answer plan.

[0457] Output: The proposed answer displayed on the user's device.

[0458] Step 7: Storing and using emotion data

[0459] Input: User emotion data analyzed by the emotion engine.

[0460] How it works: The server stores the emotional data in cloud storage. This data is used to understand the user's emotional tendencies for future response generation.

[0461] Output: Emotion data stored in cloud storage.

[0462] Step 8: Processing Natural Language Search Queries

[0463] Input: A natural language search query entered by a user through a web interface (e.g., "Show me last year's education budget discussions").

[0464] How it works: The server analyzes the search query and searches for relevant information from past meeting minutes data.

[0465] Output: Search results displayed on the user's device.

[0466] In this way, each processing step works together to automate the entire process, from creating minutes of parliamentary sessions to generating response materials for the next session of parliament.

[0467] (Application example 2)

[0468] 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."

[0469] Conventional meeting minutes creation systems simply convert the minutes into text and store them, but there is a problem in that it takes a lot of time and effort to properly prepare for the next meeting and generate draft responses. It is also difficult to create draft responses that fully reflect the user's emotions and intentions, which can lead to a decrease in user satisfaction. Furthermore, there is a demand for similar technology to be applied to other industries, and in the food delivery field in particular, there is a strong need for technology that combines review analysis and emotion recognition to improve customer satisfaction.

[0470] 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.

[0471] In this invention, the server includes means for receiving parliamentary recording data, means for analyzing the received recording data and converting the audio data into text data, means for formatting the converted text data into a predetermined format, means for saving the formatted text data, means for generating draft responses using a generative AI model based on previously saved text data, means for analyzing a user's emotional data using an emotion recognition engine and adjusting the draft responses using the generative AI model based on this, means for generating personalized messages and suggestions that reflect the user's emotional data at the time of the next order, and means for outputting the generated draft responses. This makes it possible to automatically generate appropriate draft responses and suggestions that are linked to the user's emotions.

[0472] "Parliamentary recording data" refers to audio recordings of speeches and discussions in parliament.

[0473] "Means" refers to a method or apparatus for performing a specific function or process in a patent.

[0474] "Audio data" refers to digitalized sound information acquired through a voice input device such as a microphone.

[0475] "Text data" is a form of data expressed as character information.

[0476] A "format" is a set of rules for arranging data into a particular structure or form.

[0477] A "generative AI model" is an algorithm or program that uses machine learning technology to create new information.

[0478] A "draft reply" is a proposed response to a specific question or topic in parliament or elsewhere.

[0479] An "emotion recognition engine" is a technology that analyzes and identifies human emotions from data such as voice and text.

[0480] "User" refers to a person or organization that uses the system.

[0481] A "message" is a sentence or text intended to convey specific information.

[0482] A "suggestion" is an idea or opinion submitted recommending a particular action or option.

[0483] A "system" is a collection of integrated structures or devices in which multiple elements work together to perform a specific task.

[0484] System Overview

[0485] This system applies technology for analyzing recorded conference data to generate new suggestions and messages based on users' emotional data in order to improve customer satisfaction in food delivery services. The system allows users to upload audio reviews, analyzes the audio data to recognize emotions, and provides personalized messages and suggestions for the next order. The entire system consists of a server, a terminal, and a user.

[0486] Hardware and software used

[0487] Speech Recognition: DeepSpeech, Google Speech-to-Text API

[0488] Generative AI model: GPT-4

[0489] Emotion recognition engine: IBM Watson(R) Tone Analyzer

[0490] Cloud storage: Amazon S3

[0491] Database: MySQL (registered trademark), Firebase

[0492] Web interface: React.js, Node.js

[0493] Processing Details

[0494] 1. Review upload function

[0495] After ordering from a food delivery service, users record their impressions in audio format and upload them to the app. For example, a user might upload a comment such as, "The food I ordered was delicious, but the delivery was slow."

[0496] 2. Review Analysis and Emotion Recognition

[0497] The server receives the audio files uploaded by users and stores them in Amazon S3. It then converts the audio data into text using DeepSpeech or the Google Speech-to-Text API and stores it in a MySQL database. The converted text data undergoes sentiment analysis using IBM Watson Tone Analyzer to identify the user's emotional data.

[0498] 3. Proposal generation for next order

[0499] The next time the user logs in to the app and attempts to place an order, the server generates a prompt based on the stored emotional and text data. For example, "Based on the user's review, the last order was delayed but the user was satisfied with the taste. Generate a message with highlighted menu items and a discount for the next order based on this information." This prompt is input into GPT-4, which generates an appropriate message and suggestions. The generated message and suggestions are then displayed on the user's order screen.

[0500] Specific examples

[0501] For example, if a user leaves a review about their last order saying, "The food was delicious, but the delivery was late and I was a little disappointed," the next time they order, they might see a suggestion like this: "We apologize for the delay with your last order. To ensure we can deliver your order on time, please take a look at our new Chef's Specials. Plus, we'll give you a 10% discount on your next order."

[0502] How the program is implemented

[0503] The server uses the above-mentioned methods to analyze voice data, generate emotion recognition data, generate prompts based on the stored data, and output messages and suggestions using a generative AI model. This system makes it possible to automatically generate appropriate responses and suggestions that are linked to the user's emotions.

[0504] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0505] Step 1:

[0506] A user uses a food delivery service and records an audio review after placing an order. The user opens the application, uses the recording function to record their review, and uploads the audio file (e.g., "foodie_review_2023_10_01.wav") to the app. The input is the user's voice data, which is then stored in cloud storage.

[0507] Step 2:

[0508] The server receives the audio file uploaded by the user and stores it in cloud storage (Amazon S3). Next, the server invokes a speech recognition engine (DeepSpeech or Google Speech-to-Text API) to convert the audio data into text data. In this case, the input is the audio file and the output is text data. The converted text data (e.g., "The food was delicious, but the delivery was slow") is stored in a temporary database (MySQL).

[0509] Step 3:

[0510] The server uses an emotion recognition engine (IBM Watson Tone Analyzer) to analyze the converted text data and recognize the user's emotional data. In this process, the input is text data, and the output is the user's emotional data (e.g., "dissatisfied" or "satisfied"). The analysis results are stored in cloud storage or a database.

[0511] Step 4:

[0512] The next time the user logs into the app and attempts to place a new order, the server retrieves past reviews and sentiment data. The server generates a prompt based on this data and sends it to a generative AI model (GPT-4) to generate an appropriate message or suggestion. The input is past text data and sentiment data, and the output is the generated message or suggestion. An example of a prompt is, "According to the user's review, the food was delayed on their last order, but they were satisfied with the taste. Based on this information, please generate a message including featured menu items and discounts for their next order."

[0513] Step 5:

[0514] The generative AI model (GPT-4) generates new suggestions and messages based on the prompt. For example, "We apologize for the delay in your last order. As a special favor, we invite you to try our new menu item. We'll give you a 10% discount on your next order." The input is the prompt, and the output is a personalized message.

[0515] Step 6:

[0516] The server displays the generated message on the user's order screen. The user can then place a new order based on the displayed message and suggestions. The input is the generated message, and the output is the user's display screen. This improves the user experience and increases ordering satisfaction.

[0517] 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.

[0518] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (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.

[0519] 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.

[0520] [Second embodiment]

[0521] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0522] 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.

[0523] 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).

[0524] 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.

[0525] 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.

[0526] 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).

[0527] 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. 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.

[0528] 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.

[0529] 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.

[0530] 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.

[0531] In the smart glasses 214, 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.

[0532] 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."

[0533] The system of the present invention uses a generative AI model to create minutes of parliamentary sessions and automatically generate materials for the next session (such as draft answers) based on that data. Below, we will explain how this system is implemented in concrete terms.

[0534] Acquisition of parliamentary recording data

[0535] The user uploads a new session recording file to the server. For example, after a session ends, the user uploads the recording data (e.g., "session_2023_10.wav") via the web interface. The file is received by the server and stored in cloud storage.

[0536] Data analysis and minutes preparation

[0537] The server analyzes the uploaded recording file. A speech recognition engine is used for this analysis, converting the audio data into text data. Speech recognition technologies such as DeepSpeech and Google Speech-to-Text API are used. The converted text data is split into segments for each speaker and appropriately formatted using speaker identification technology. It is then converted into a format (e.g., JSON) that is easy for the generative AI model to understand. Here is an example:

[0538] json

[0539] {

[0540] "session_id": "2023_10",

[0541] "timestamp": "2023-10-15",

[0542] "content": [

[0543] {

[0544] "speaker": "Rep. A",

[0545] "statement": "I'd like to ask about this year's budget."

[0546] },

[0547] {

[0548] "speaker": "Mayor B",

[0549] "statement": "The reason for the budget cuts is..."

[0550] }

[0551] ]

[0552] }

[0553] Storage of past meeting minutes data

[0554] The server saves the formatted minutes data in cloud storage. When saving, metadata (e.g., session ID, date) is also saved to ensure association with past minutes data.

[0555] Generate next parliamentary response

[0556] The user inputs the topic and questions for the next assembly meeting on the terminal. For example, the user might input, "I plan to ask a question about the budget at the next assembly meeting." The server then searches past minutes data based on this input and extracts relevant information.

[0557] The server uses a generative AI model (e.g., GPT-4) to generate new draft answers based on the extracted information. The generated draft answers are displayed on the device for the user to review and edit. For example, the following draft answers may be generated:

[0558] json

[0559] {

[0560] "suggestions": [

[0561] {

[0562] "topic": "budget",

[0563] "answer": "The budget cuts made last year were made to improve efficiency and were implemented in a way that minimized the impact on the educational environment."

[0564] }

[0565] ]

[0566] }

[0567] Natural language search function

[0568] A user can search the database of past meeting minutes using a natural language search query. For example, they can type "Show me last year's education budget discussion." The server analyzes this natural language query and searches for relevant meeting minutes data. As a result, the following search results are displayed on the user's device:

[0569] text

[0570] Meeting on October 15, 2022:

[0571] Councilman A: "Please explain last year's education budget cuts."

[0572] Mayor B: "The reason behind the budget cuts is..."

[0573] Parliamentary recording and video recording equipment

[0574] Users can also use audio and video recording equipment specifically designed for parliamentary use. This equipment is designed to record high-quality audio and video recordings of parliamentary speeches and debates. The audio and video recordings are then uploaded to a server for later analysis and storage.

[0575] This system allows local government employees to carry out their work efficiently and significantly reduces the time it takes to prepare assembly response proposals. In addition, by utilizing past meeting minutes data, it is possible to prepare consistent response proposals.

[0576] The processing flow will be explained below.

[0577] Step 1:

[0578] The user uploads the parliamentary recording (e.g., "session_2023_10.wav") to the server using the interface. The user selects the recording using a web browser and clicks the "Upload" button.

[0579] Step 2:

[0580] The server receives the uploaded recording data. After receiving it, the file is stored in cloud storage, where the file's metadata (e.g., session ID, upload date and time) is also recorded.

[0581] Step 3:

[0582] The server analyzes the received recording data, activates a speech recognition engine, and converts the audio data into text data, which is then stored in a temporary database.

[0583] Step 4:

[0584] The server extracts the text data obtained from the speech recognition engine and uses speaker identification technology to segment it into segments for each speaker, identifying speakers based on, for example, voice patterns or time of day.

[0585] Step 5:

[0586] The server formats the text data of the identified speaker into a format that is easy for the AI ​​model to understand (e.g., JSON format). The formatted data is then stored in cloud storage. The following is an example of the data:

[0587] json

[0588] {

[0589] "session_id": "2023_10",

[0590] "timestamp": "2023-10-15",

[0591] "content": [

[0592] {

[0593] "speaker": "Rep. A",

[0594] "statement": "I'd like to ask about this year's budget."

[0595] },

[0596] {

[0597] "speaker": "Mayor B",

[0598] "statement": "The reason for the budget cuts is..."

[0599] }

[0600] ]

[0601] }

[0602] Step 6:

[0603] The user enters the topic and question for the next session of the Diet on the device. For example, they enter, "I plan to ask a question about the budget at the next session of the Diet." The input is done through a web interface.

[0604] Step 7:

[0605] The server searches for past minutes data based on the input topic, and efficiently extracts related minutes data using metadata.

[0606] Step 8:

[0607] The server uses a generative AI model (e.g., GPT-4) to generate a new draft response based on the extracted minutes data. Below is an example of a generated draft response:

[0608] json

[0609] {

[0610] "suggestions": [

[0611] {

[0612] "topic": "budget",

[0613] "answer": "The budget cuts made last year were made to improve efficiency and were implemented in a way that minimized the impact on the educational environment."

[0614] }

[0615] ]

[0616] }

[0617] Step 9:

[0618] The generated draft answer is displayed on the device, where the user can review it and edit it if necessary. The final draft answer can also be saved in a different format or printed.

[0619] Step 10:

[0620] A user enters a natural language search query, for example, "Show me last year's education budget discussions."

[0621] Step 11:

[0622] The server analyzes the entered search query and searches the database for relevant meeting minutes data, using natural language processing techniques.

[0623] Step 12:

[0624] The search results are displayed on the device. For example, the following search results are displayed on the user's screen:

[0625] text

[0626] Meeting on October 15, 2022:

[0627] Councilman A: "Please explain last year's education budget cuts."

[0628] Mayor B: "The reason behind the budget cuts is..."

[0629] Example 1

[0630] 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."

[0631] Conventional methods for managing parliamentary recording data require a great deal of time and effort for manual minutes creation and data search, resulting in low productivity. It is also difficult to quickly propose consistent content for the next parliamentary session. Therefore, there is a need for a system that can automatically and efficiently create minutes and provide draft answers using generative AI models.

[0632] 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.

[0633] In this invention, the server includes means for receiving recorded data, means for analyzing the received recorded data and converting the audio data into text data, means for formatting the converted text data into a predetermined format, means for saving the formatted text data, means for generating a draft answer using a generative AI model based on previously saved text data, means for outputting the generated draft answer, and means for displaying the output draft answer on a terminal for user input. This makes it possible to automatically and efficiently create minutes and quickly propose consistent draft answers using the generative AI model.

[0634] "Recorded data" refers to data that records in audio format statements and discussions at meetings such as parliaments.

[0635] "Means for receiving" refers to the function or device that allows the server to import the recorded data uploaded by the user.

[0636] "Means for analyzing" refers to the technology or method for analyzing the received recorded data and converting the audio data into text data.

[0637] "Audio data" refers to audio signals stored as recorded data.

[0638] "Text data" is data of character strings that are obtained by analyzing voice data and converting it into linguistic information.

[0639] The "predetermined format" refers to the form or structure used to format the generated text data to ensure consistency and ease of understanding.

[0640] "Storage means" refers to a function or system for storing and managing formatted text data.

[0641] A "generative AI model" refers to a model that uses artificial intelligence to generate responses or suggestions based on specific input data.

[0642] A "draft answer" is a prepared answer that a member of parliament or other relevant person can use to appropriately respond to a question asked in a parliament or other such setting.

[0643] "Means of output" refers to the functions and systems for presenting the generated text data and draft answers to the user.

[0644] A "user input terminal" is a device used by a user to enter data and check results.

[0645] A "natural language search query" refers to a question that a user enters in natural language to search past data.

[0646] "Query-based search means" refers to techniques or methods for retrieving relevant past data from an input natural language search query.

[0647] "Means for outputting search results" refers to the functions and systems for displaying search results to the user.

[0648] "Recording device" refers to a device or equipment for acquiring audio and video data.

[0649] This invention is a system that uses generative AI models to efficiently create parliamentary minutes and automatically generate materials for the next parliamentary session. The system starts with the user uploading audio data, then analyzes the audio data, converts it to text data, stores the data, searches it, and generates draft responses.

[0650] Acquisition of parliamentary recording data

[0651] A user uploads the recorded data to the server via a web interface. For example, after a session ends, the user selects the recorded data (e.g., "session_2023_10.wav") and clicks the upload button. The server receives the data as an HTTP POST request and saves it to cloud storage (e.g., Amazon S3). The server notifies the user when the file has been saved.

[0652] Data analysis and minutes preparation

[0653] The server passes the received recording data to a speech recognition engine, which converts the audio data into text data. This analysis is performed using, for example, the Google Speech-to-Text API. The audio data is converted into a byte stream and sent via the API. The returned text data is split into segments for each speaker using speaker identification technology and formatted into a JSON format that is easy for the generative AI model (e.g., GPT-4) to understand.

[0654] For example, the following JSON data is generated:

[0655] json

[0656] {

[0657] "session_id": "2023_10",

[0658] "timestamp": "2023-10-15",

[0659] "content": [

[0660] {

[0661] "speaker": "Rep. A",

[0662] "statement": "I'd like to ask about this year's budget."

[0663] },

[0664] {

[0665] "speaker": "Mayor B",

[0666] "statement": "The reason for the budget cuts is..."

[0667] }

[0668] ]

[0669] }

[0670] Storage of past meeting minutes data

[0671] The server saves the formatted JSON data in cloud storage. This allows it to be associated with past meeting minutes data. When saving, metadata (e.g., session ID, date) is also saved.

[0672] Generate next parliamentary response

[0673] The user uses an input form on their device to input the topic and questions for the next session. For example, they might input, "I plan to ask a question about the budget at the next session." The server receives this input and searches past minutes data to extract relevant information. It then uses a generative AI model (e.g., GPT-4) to generate new draft responses. An example of this prompt is, "Imagine a question about 'budget cuts' at the next session. Please generate a relevant draft response based on past minutes data." The generated draft response is displayed on the user's device.

[0674] Natural language search function

[0675] Users can search past meeting minutes data by entering natural language search queries into their devices. For example, they can enter "Show me last year's education budget discussions." The server analyzes this input using a natural language processing (NLP) module and searches for relevant data. Search results are displayed on the user's device.

[0676] You'll see specific search results like this:

[0677] text

[0678] Meeting on October 15, 2022:

[0679] Councilman A: "Please explain last year's education budget cuts."

[0680] Mayor B: "The reason behind the budget cuts is..."

[0681] Parliamentary recording and video recording equipment

[0682] Users operate dedicated audio and video recording equipment for council meetings to record speeches and discussions in high quality. The recorded data is later uploaded to a server for analysis and storage. This system enables local government officials to carry out their work efficiently and prepare consistent responses.

[0683] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0684] Step 1: Upload your recordings

[0685] The user uploads the recording (e.g., "session_2023_10.wav") to the server via a web interface. Input: The recording file of the parliament. Output: The uploaded recording file is saved in cloud storage. Specific behavior: The user opens a file selection dialog in a web browser, selects the recording data from the local file system, and presses the upload button. The server receives the HTTP POST request and saves the file in cloud storage.

[0686] Step 2: Analyzing the audio data

[0687] The server retrieves the recorded data from cloud storage and sends it to a speech recognition engine (e.g., Google Speech-to-Text API). Input: The saved recording file. Output: The speech data converted to text data. Specific operation: The server converts the speech data into a byte stream, sends it to the recognition engine, and receives the returned text data.

[0688] Step 3: Formatting the text data

[0689] The server analyzes the acquired text data and uses speaker identification technology to split and format it for each speaker. Input: Text data returned from the speech recognition engine. Output: Meeting minutes data in JSON format. Specific operation: The server analyzes the text data, splits it for each speaker, and generates formatted JSON data.

[0690] Step 4: Save your data

[0691] The server saves the formatted JSON data in cloud storage. Input: Formatted text data (JSON format). Output: Meeting minutes data saved in cloud storage. Specific operation: The server saves the JSON data in the specified directory, along with the metadata.

[0692] Step 5: Generate draft answers for the next session

[0693] The user inputs the topic and question for the next session of the Diet (e.g., "I plan to ask a question about the budget at the next session of the Diet") into the device. Input: The topic and question for the next session of the Diet. Output: Generated draft answers. Specific operation: The server searches past minutes data based on the input topic and question, generates new draft answers using a generative AI model (e.g., GPT-4), and displays them on the device.

[0694] Step 6: Natural Language Search

[0695] A user enters a natural language search query (e.g., "Show me last year's education budget discussions") into a device. Input: Natural language search query. Output: Relevant meeting minutes data. Specific behavior: The server uses a natural language processing (NLP) module to parse the search query, search for relevant data, and display the search results on the device.

[0696] Step 7: Operating the recording equipment

[0697] The user operates audio and video recording equipment specifically designed for the assembly to capture high-quality audio and video data. Input: Data captured by the audio and video recording equipment. Output: The audio and video data is uploaded to the server and stored in cloud storage. Specific operation: The user presses the start recording button on the audio and video recording equipment, which records the entire assembly meeting and then uploads the data to the server.

[0698] (Application example 1)

[0699] 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."

[0700] In logistics centers, creating minutes of meetings and conferences is a time-consuming and labor-intensive process, and labor shortages can lead to inefficiencies. It is also difficult to quickly retrieve information related to the next meeting or task by searching past meeting records. The present invention aims to solve these problems by providing a system that automatically converts meeting recording data into text and generates materials and plans for the next meeting based on a generative AI model.

[0701] 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.

[0702] In this invention, the server includes means for receiving recorded assembly data, means for analyzing the received recorded data and converting the audio data into text data, means for formatting the converted text data into a predetermined format, means for saving the formatted text data, means for generating draft responses using a generative AI model based on previously saved text data, means for inputting the next meeting theme and tasks, means for extracting related information and generating new meeting materials and draft plans, and means for outputting the generated draft responses and meeting materials. This automates the creation of meeting minutes within the logistics center, enabling the efficiency of preparations for the next meeting and tasks.

[0703] "Parliamentary recording data" refers to data files containing recordings of speeches and discussions held in parliaments or meetings.

[0704] "Means for receiving" refers to the function for obtaining recorded data from outside and importing it into the server.

[0705] "Means for analyzing and converting voice data into text data" refers to a function for converting recorded voice data into text information using voice recognition technology.

[0706] "Means for formatting into a specified format" refers to a function for arranging converted text data according to certain rules and formats.

[0707] "Means of storage" refers to the function of storing formatted text data in a database, cloud storage, etc. for future reference or search.

[0708] "Means for generating draft responses using generative AI models" refers to a function that uses AI technology based on past data to automatically create draft responses to be used in the next parliament or meeting.

[0709] "Means for inputting the next meeting topic and tasks" refers to an interface that allows the user to provide the server with information and questions about the next meeting.

[0710] "Means for extracting relevant information and generating new materials and plans" refers to the function of extracting necessary information from past data and using it to create new materials and plans.

[0711] "Means for outputting" refers to a function for providing the generated draft answers and meeting materials to the user.

[0712] "Means for accepting natural language search queries" refers to an interface that allows users to enter search terms in everyday language.

[0713] "Means for searching previously stored text data based on a search query" refers to a function for extracting data that matches the user's input from a database.

[0714] "Means for outputting search results" refers to a function for providing the searched content to the user.

[0715] "Means for additionally analyzing recorded data" refers to a function for further analyzing recorded data within a logistics facility and converting it into text data, etc.

[0716] "Function to record meeting recording data" refers to the function to collect the contents of meetings and assemblies in audio format.

[0717] "The function of recording recorded data" refers to the function of collecting the contents of meetings and assemblies in video format.

[0718] "Recording and video recording equipment" refers to devices and equipment used to record the contents of meetings and assemblies as audio or video.

[0719] MODE FOR CARRYING OUT THE INVENTION

[0720] This system automatically analyzes recorded meeting data at a logistics center, converts it into text data, and saves it. It can also automatically generate materials and plans for the next meeting using a generative AI model. It also has the ability to search past meeting records using natural language search queries and provide the results to users.

[0721] System Overview

[0722] The system consists of the following main components:

[0723] 1. Means of receiving parliamentary recordings

[0724] 2. Means of converting audio data into text data

[0725] 3. A means of formatting text data into a specified format

[0726] 4. Means of saving text data

[0727] 5. How to generate draft responses using generative AI models

[0728] 6. A way to enter the next meeting topic or task

[0729] 7. A means to extract relevant information and generate new meeting materials and plans

[0730] 8. Means for outputting the generated response plan and meeting materials

[0731] 9. A way to accept natural language search queries

[0732] 10. A way to search past data based on a search query

[0733] 11. How to output search results

[0734] 12. Additional means of analyzing recording data

[0735] 13. Function to record conference recording data

[0736] 14. Audio and video recording devices with the function of recording video data

[0737] System Operation

[0738] The server receives the parliamentary recording data, analyzes it, and converts the audio data into text data. This speech recognition uses technologies such as DeepSpeech and Google Speech-to-Text API. The converted text data is formatted and stored in cloud storage (e.g., AWS S3).

[0739] Users can input the topic or task for the next meeting into their device. For example, they can input a prompt such as, "Please create a document to discuss the logistics situation at the next meeting." The server then searches past text data and extracts relevant information. Based on this extracted information, a generative AI model (e.g., GPT-4) is used to generate new meeting materials and draft plans. The generated materials and draft responses are then displayed on the user's device.

[0740] Additionally, users can search for past meeting records using natural language search queries, such as a prompt such as "Show me meeting records from last year regarding logistics status." The server analyzes the query, searches for relevant data, and provides the results to the user.

[0741] Hardware and software used

[0742] The system uses the following hardware and software:

[0743] Server: Amazon Web Services (AWS) EC2

[0744] Cloud storage: AWS S3

[0745] Speech recognition engine: DeepSpeech, Google Speech-to-Text API

[0746] Generative AI model: OpenAI GPT-4

[0747] Natural language processing libraries: SpaCy, NLTK

[0748] Specific examples

[0749] For example, suppose a logistics center manager holds a weekly meeting and records it. The recording is then uploaded to a server, which analyzes the data and creates meeting minutes. For next week's meeting, if a user types into their device, "Please generate this week's plan based on last week's minutes," the server will search past data and automatically generate a new plan using a generative AI model. This can improve work efficiency.

[0750] Examples of prompts include "Please prepare materials to talk about the logistics situation at the next meeting" and "Show me the minutes of last year's meeting about the logistics situation."

[0751] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0752] Step 1:

[0753] A user acquires parliamentary recording data using audio and video recording equipment and uploads the data to the server. In this example, the recording file (e.g., "meeting_2023_10.wav") is sent to the server via a web interface. The input is an audio file, and the output is an audio file stored in cloud storage.

[0754] Step 2:

[0755] The server analyzes the received recording data and converts the audio data into text data. This process uses a speech recognition engine (DeepSpeech or Google Speech-to-Text API). The input is an audio file, and the output is the converted text data.

[0756] Step 3:

[0757] The server formats the converted text data into a specified format. Specifically, it uses speaker identification technology to split the text data for each speaker and converts it into JSON format. The input is raw text data, and the output is formatted text data.

[0758] Step 4:

[0759] The server saves the formatted text data in cloud storage (AWS S3), along with metadata (e.g., session ID, date). The input is the formatted text data, and the output is the data saved in cloud storage.

[0760] Step 5:

[0761] The user inputs the next meeting topic and task into the terminal. For example, the user might input, "Please prepare materials to talk about the logistics situation at the next meeting." The input is a prompt text, and the output is sending the prompt text to the server.

[0762] Step 6:

[0763] The server searches for information related to the prompt text entered by the user based on previously saved text data. The input is the prompt text and past text data, and the output is the searched related information.

[0764] Step 7:

[0765] Based on the related information retrieved by the server, a generative AI model (GPT-4) is used to generate meeting materials and plan proposals. The input is related information, and the output is new meeting materials and plan proposals.

[0766] Step 8:

[0767] The server outputs the generated meeting materials and plans to the terminals to provide them to the users. The input is the new meeting materials and plans, and the output is the materials and plans displayed on the user's terminal.

[0768] Step 9:

[0769] A user inputs a search query in natural language into a terminal. For example, "Show me the minutes of the meeting about the logistics situation last year." The input is a search query in natural language, and the output is a search query sent to the server.

[0770] Step 10:

[0771] The server searches past text data based on the search query. Here, it references a database of past meeting minutes and extracts relevant data. The input is the search query and past text data, and the output is the search result data.

[0772] Step 11:

[0773] The server outputs the search results to the terminal, allowing the user to view the results. The input is the search result data, and the output is the search result displayed on the user's terminal.

[0774] 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.

[0775] The system of the present invention uses a generative AI model to create minutes of parliamentary sessions and automatically generates the next parliamentary session materials (such as draft parliamentary responses) based on that data. Furthermore, by combining it with an emotion engine that recognizes the user's emotions, it is possible to generate more appropriate draft responses that reflect the user's intentions and emotions. Below, we will explain in detail how this system is implemented.

[0776] Acquisition of parliamentary recording data

[0777] The user uploads the parliamentary recording (e.g., "session_2023_10.wav") to the server using the interface. The user selects the recording via a web browser and clicks the "Upload" button. The file is received by the server and stored in cloud storage.

[0778] Data analysis and minutes preparation

[0779] The server analyzes the uploaded recording data, activates a speech recognition engine, and converts the audio data into text data. The speech recognition technology used is DeepSpeech or the Google Speech-to-Text API, for example. The converted text data is then stored in a temporary database.

[0780] Speaker Identification and Data Shaping

[0781] The server extracts the text data obtained from the speech recognition engine and divides it into segments for each speaker using speaker identification technology. For example, speakers can be identified based on voice patterns or time of day. The server then formats the text data of the identified speaker into a format (e.g., JSON) that is easy for the generative AI model to understand. The formatted data is then saved in cloud storage. The following is an example of the data:

[0782] json

[0783] {

[0784] "session_id": "2023_10",

[0785] "timestamp": "2023-10-15",

[0786] "content": [

[0787] {

[0788] "speaker": "Rep. A",

[0789] "statement": "I'd like to ask about this year's budget."

[0790] },

[0791] {

[0792] "speaker": "Mayor B",

[0793] "statement": "The reason for the budget cuts is..."

[0794] }

[0795] ]

[0796] }

[0797] Generating Answer Plans and Using an Emotion Engine

[0798] The user inputs the topic and questions for the next session of the Diet on the terminal. For example, the user might input, "I plan to ask a question about the budget at the next session of the Diet." The input is done via a web interface. The server searches past minutes data based on this input and extracts relevant information.

[0799] The server uses a generative AI model (e.g., GPT-4) to generate new response proposals based on the extracted information. Furthermore, an emotion engine recognizes the user's emotional data and adjusts the response proposals based on that data. For example, if the user is recognized as "nervous," the response proposals will include expressions that convey a more reassuring feeling.

[0800] json

[0801] {

[0802] "suggestions": [

[0803] {

[0804] "topic": "budget",

[0805] "answer": "The budget cuts made last year were made to improve efficiency and we have taken measures to minimize the impact on the educational environment. Please rest assured."

[0806] }

[0807] ]

[0808] }

[0809] Storing and using emotional data

[0810] The server stores the user's emotional data recognized by the emotion engine in cloud storage. The stored emotional data is used to generate future parliamentary response drafts. This results in more consistent response drafts and increases user confidence.

[0811] Natural language search function

[0812] A user can search the database of past meeting minutes using a natural language search query. For example, they can type "Show me last year's education budget discussion." The server analyzes this natural language query and searches for relevant meeting minutes data. As a result, the following search results are displayed on the user's device:

[0813] text

[0814] Meeting on October 15, 2022:

[0815] Councilman A: "Please explain last year's education budget cuts."

[0816] Mayor B: "The reason behind the budget cuts is..."

[0817] Parliamentary recording and video recording equipment

[0818] Users can record high-quality audio and video of parliamentary speeches and debates using dedicated parliamentary audio and video recording equipment, which will then be uploaded to a server for later analysis and storage.

[0819] This system allows local government employees to carry out their work efficiently and significantly reduces the time it takes to prepare assembly response drafts. Furthermore, by taking into consideration the user's feelings, it is possible to provide more reliable response drafts. In addition, by utilizing past minutes data, it is possible to prepare consistent response drafts.

[0820] The processing flow will be explained below.

[0821] Step 1:

[0822] The user uploads the parliamentary recording (e.g., "session_2023_10.wav") to the server using the interface. The user selects the recording via a web browser and clicks the "Upload" button.

[0823] Step 2:

[0824] The server receives the uploaded recording data. After receiving it, the file is stored in cloud storage, where the file's metadata (e.g., session ID, upload date and time) is also recorded.

[0825] Step 3:

[0826] The server analyzes the received recording, invokes a speech recognition engine (e.g., DeepSpeech or Google Speech-to-Text API), and converts the audio data into text data, which is then stored in a temporary database.

[0827] Step 4:

[0828] The server extracts the text data obtained from the speech recognition engine and uses speaker identification technology to segment it into segments for each speaker, identifying speakers based on, for example, voice patterns or time of day.

[0829] Step 5:

[0830] The server formats the text data of the identified speaker into a format that is easy for the AI ​​model to understand (e.g., JSON format). The formatted data is then stored in cloud storage. The following is an example of the data:

[0831] json

[0832] {

[0833] "session_id": "2023_10",

[0834] "timestamp": "2023-10-15",

[0835] "content": [

[0836] {

[0837] "speaker": "Rep. A",

[0838] "statement": "I'd like to ask about this year's budget."

[0839] },

[0840] {

[0841] "speaker": "Mayor B",

[0842] "statement": "The reason for the budget cuts is..."

[0843] }

[0844] ]

[0845] }

[0846] Step 6:

[0847] The user inputs the topic and questions for the next council meeting on the terminal. For example, the user inputs, "I plan to ask a question about the budget at the next council meeting." The input is done via a web interface.

[0848] Step 7:

[0849] The server searches for past minutes data based on the input topic, and efficiently extracts related minutes data using metadata.

[0850] Step 8:

[0851] The server uses a generative AI model (e.g., GPT-4) to generate a new draft response based on the extracted minutes data. Below is an example of a generated draft response:

[0852] json

[0853] {

[0854] "suggestions": [

[0855] {

[0856] "topic": "budget",

[0857] "answer": "The budget cuts made last year were made to improve efficiency and were implemented in a way that minimized the impact on the educational environment."

[0858] }

[0859] ]

[0860] }

[0861] Step 9:

[0862] The server uses the emotion engine to recognize the user's emotion, for example, the emotion engine recognizes that the user is "nervous."

[0863] Step 10:

[0864] The server reflects the emotional data it recognizes in the generated answer proposals. For example, if the server recognizes that the user is "nervous," the answer proposals include expressions that make the user feel more at ease.

[0865] Step 11:

[0866] The generated draft answer is displayed on the device, where the user can review it and edit it if necessary. The final draft answer can also be saved in a different format or printed.

[0867] Step 12:

[0868] A user enters a natural language search query, for example, "Show me last year's education budget discussions."

[0869] Step 13:

[0870] The server analyzes the entered search query and searches the database for relevant meeting minutes data, using natural language processing techniques.

[0871] Step 14:

[0872] The search results are displayed on the device. For example, the following search results are displayed on the user's screen:

[0873] text

[0874] Meeting on October 15, 2022:

[0875] Councilman A: "Please explain last year's education budget cuts."

[0876] Mayor B: "The reason behind the budget cuts is..."

[0877] Example 2

[0878] 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."

[0879] The task of creating minutes of assembly meetings and response materials for the next assembly meeting requires time and effort, placing a heavy burden on staff. Furthermore, creating response materials requires appropriately reflecting the content of past discussions while also taking into account the user's emotions and intentions, but there is no efficient way to do this. In conventional technologies, creating minutes and generating response materials is often done manually, resulting in reduced work efficiency and inconsistent quality. This invention is intended to solve these problems.

[0880] The specific processing by the specific processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for receiving parliamentary recording data, means for analyzing the received recording data and converting the audio data into text data, means for formatting the converted text data into a predetermined data format, means for generating a draft response using a generative AI model based on previously saved text data, means for reflecting the user's emotions in the generated draft response using an emotion engine, and means for outputting the generated draft response. This automates the creation of minutes and the generation of response materials, improving work efficiency and ensuring consistent quality. Furthermore, because response materials are generated taking the user's emotions into consideration, more appropriate and reliable materials can be created.

[0881] "Parliamentary recording data" refers to audio data of speeches and discussions in parliament.

[0882] A "generative AI model" refers to an artificial intelligence model that generates new outputs based on input data.

[0883] "Emotion engine" refers to a program or function that recognizes a user's emotions and analyzes the emotion data.

[0884] "Draft answers" refers to prepared answers to questions asked in parliament.

[0885] "Converted text data" refers to character data converted from voice data using voice recognition technology.

[0886] "Data format" refers to the format or standard for organizing and arranging data in a certain structure.

[0887] A "natural language search query" refers to a search request made to a database using natural language.

[0888] "Sound recording equipment" means devices for recording sound and video.

[0889] "Cloud storage" refers to an online storage service that stores and manages data via the Internet.

[0890] A "server" refers to a computer system that provides services and data to other computers over a network.

[0891] "Terminal" refers to a device used by a user for input and display.

[0892] "Text data" refers to data expressed as character information.

[0893] The system of the present invention utilizes a generative AI model and an emotion engine to create minutes of parliamentary meetings and automatically generate materials for responses to questions at the next parliamentary meeting. A detailed embodiment of this system will be described.

[0894] Overall system overview

[0895] This system analyzes recordings of parliamentary sessions, converts the audio data into text data, and uses that text data to automatically generate draft responses for the next session. It can also recognize the user's emotions and generate draft responses that reflect those emotions. This system consists of three main components: a server, a terminal, and a user.

[0896] Hardware and Software Configuration

[0897] The server receives the recorded data and analyzes the audio data. It is equipped with a high-performance processor and a large amount of memory. The speech recognition engine uses common speech recognition software such as DeepSpeech or Google Speech-to-Text API. The server also converts the analyzed text data into a specified data format and stores it in a temporary database.

[0898] The terminal provides an interface for users to input the topic and questions for the next session. This terminal is a standard PC or tablet and is connected to the server via a web browser. Users can use the terminal to upload recording data as well as check the generated response plans.

[0899] Users upload the recorded data of assembly meetings to the server using their devices. They input the topic and questions of the next assembly meeting, and check and use the proposed answers provided by the server. Users can also input search queries to search for past minutes data.

[0900] Processing flow and specific examples

[0901] The system begins by a user uploading a recording of a parliamentary session. The recording (e.g., "session_2023_10.wav") is then stored in cloud storage by the server. The server then activates a speech recognition engine, analyzes the recording, and converts the audio data into text data. This text data is then split by speaker using speaker identification technology and formatted into a format that is easy for the generative AI model to understand.

[0902] For example, a statement such as "I'd like to ask about this year's budget" is linked to "Council Member A," and a statement such as "The reason for the budget cuts is..." is linked to "Mayor B." This results in the data being formatted as text like this:

[0903] Session ID: 2023_10

[0904] Timestamp: 2023-10-15

[0905] Contents:

[0906] Councilman A: "I'd like to ask about this year's budget."

[0907] Mayor B: 'The reason for the budget cuts is...'

[0908] "

[0909] Next, the user inputs the next session's theme and question on the device. For example, they might input, "I plan to ask a question about the budget at the next session." The server extracts relevant information from past minutes data and uses this information to input prompts into the generative AI model to generate new draft responses. The emotion engine also recognizes the user's emotions, and if it recognizes that the user is "nervous," the generated draft responses will include expressions that convey a sense of security.

[0910] Examples of prompts are:

[0911] "Parliamentary Data:

[0912] Session ID: 2023_10

[0913] Timestamp: 2023-10-15

[0914] Contents:

[0915] Councilman A: "I'd like to ask about this year's budget."

[0916] Mayor B: 'The reason for the budget cuts is...'

[0917] Next Congress Theme and User Sentiments:

[0918] Theme: 'Budget'

[0919] Question: I plan to ask a question about the budget at the next Congress. I would like to know specifically why the budget cuts were made.

[0920] Emotion: 'Nervous'

[0921] Generated answer plan:

[0922] Topic: 'Budget'

[0923] Response: 'The budget cuts made last year were made to improve efficiency and we have taken measures to minimize the impact on the educational environment. Please rest assured.'

[0924] "

[0925] The server saves the generated answer plan in cloud storage and displays it on the user's device. The user's emotional data is also saved in cloud storage and used to generate answer plans from the next time onwards, making it possible to create consistent answer plans.

[0926] The above is a specific embodiment of the system. This system automates the creation of assembly minutes and the generation of response materials for the next assembly meeting, and can provide high-quality materials that take user emotions into consideration.

[0927] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0928] Step 1: Obtaining Congressional Recording Data

[0929] Input: The user uploads a recording file (e.g., "session_2023_10.wav") through the designated interface.

[0930] How it works: A user uses a web browser to select a recording and presses the "upload" button.

[0931] Output: The server saves this recording to cloud storage and records the file's metadata (e.g., file name, upload date and time).

[0932] Step 2: Analyze the data and create a transcript

[0933] Input: Recording data files stored in cloud storage.

[0934] How it works: The server launches a speech recognition engine (e.g., DeepSpeech or Google Speech-to-Text API) and reads the recording. The server converts the audio data into text. This conversion process converts what is being said into text format.

[0935] Output: The converted text data is saved in a temporary database.

[0936] Step 3: Speaker identification and data conditioning

[0937] Input: Text data stored in a temporary database.

[0938] How it works: The server uses speaker identification technology to identify each speaker, using characteristics such as voice patterns and time of day. It then organizes the text data for each identified speaker and formats it into a data format (e.g., JSON) that is easy for the generative AI model to understand.

[0939] Output: The formatted text data is saved to cloud storage.

[0940] Step 4: Enter the theme and questions for the next assembly meeting

[0941] Input: The user enters the topic or question for the next session through a web interface (e.g., "I have a question about the budget for the next session").

[0942] How it works: The user uses a terminal to enter the required information and send it to the server.

[0943] Output: The server receives this input data and prepares it for the next processing step.

[0944] Step 5: Generate a response and use the emotion engine

[0945] Input: User questions and past meeting minutes data.

[0946] How it works: The server extracts relevant information from past meeting minutes data and inputs prompt sentences into a generative AI model. A generative AI model (e.g., GPT-4) is used to generate new response proposals. The emotion engine then analyzes the user's emotion data and adjusts the response proposals based on the recognized emotion. For example, "If the user is nervous, add expressions that will make them feel at ease."

[0947] Output: A response plan reflecting the sentiment is generated and saved in cloud storage.

[0948] Step 6: Outputting the generated answer plan

[0949] Input: Response plan data stored on the server.

[0950] Operation: The server sends the generated answer plan to the user's terminal, where the user can view the answer plan.

[0951] Output: The proposed answer displayed on the user's device.

[0952] Step 7: Storing and using emotion data

[0953] Input: User emotion data analyzed by the emotion engine.

[0954] How it works: The server stores the emotional data in cloud storage. This data is used to understand the user's emotional tendencies for future response generation.

[0955] Output: Emotion data stored in cloud storage.

[0956] Step 8: Processing Natural Language Search Queries

[0957] Input: A natural language search query entered by a user through a web interface (e.g., "Show me last year's education budget discussions").

[0958] How it works: The server analyzes the search query and searches for relevant information from past meeting minutes data.

[0959] Output: Search results displayed on the user's device.

[0960] In this way, each processing step works together to automate the entire process, from creating minutes of parliamentary sessions to generating response materials for the next session of parliament.

[0961] (Application example 2)

[0962] 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."

[0963] Conventional meeting minutes creation systems simply convert the minutes into text and store them, but there is a problem in that it takes a lot of time and effort to properly prepare for the next meeting and generate draft responses. It is also difficult to create draft responses that fully reflect the user's emotions and intentions, which can lead to a decrease in user satisfaction. Furthermore, there is a demand for similar technology to be applied to other industries, and in the food delivery field in particular, there is a strong need for technology that combines review analysis and emotion recognition to improve customer satisfaction.

[0964] 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.

[0965] In this invention, the server includes means for receiving parliamentary recording data, means for analyzing the received recording data and converting the audio data into text data, means for formatting the converted text data into a predetermined format, means for saving the formatted text data, means for generating draft responses using a generative AI model based on previously saved text data, means for analyzing a user's emotional data using an emotion recognition engine and adjusting the draft responses using the generative AI model based on this, means for generating personalized messages and suggestions that reflect the user's emotional data at the time of the next order, and means for outputting the generated draft responses. This makes it possible to automatically generate appropriate draft responses and suggestions that are linked to the user's emotions.

[0966] "Parliamentary recording data" refers to audio recordings of speeches and discussions in parliament.

[0967] "Means" refers to a method or apparatus for performing a specific function or process in a patent.

[0968] "Audio data" refers to digitalized sound information acquired through a voice input device such as a microphone.

[0969] "Text data" is a form of data expressed as character information.

[0970] A "format" is a set of rules for arranging data into a particular structure or form.

[0971] A "generative AI model" is an algorithm or program that uses machine learning technology to create new information.

[0972] A "draft reply" is a proposed response to a specific question or topic in parliament or elsewhere.

[0973] An "emotion recognition engine" is a technology that analyzes and identifies human emotions from data such as voice and text.

[0974] "User" refers to a person or organization that uses the system.

[0975] A "message" is a sentence or text intended to convey specific information.

[0976] A "suggestion" is an idea or opinion submitted recommending a particular action or option.

[0977] A "system" is a collection of integrated structures or devices in which multiple elements work together to perform a specific task.

[0978] System Overview

[0979] This system applies technology for analyzing recorded conference data to generate new suggestions and messages based on users' emotional data in order to improve customer satisfaction in food delivery services. The system allows users to upload audio reviews, analyzes the audio data to recognize emotions, and provides personalized messages and suggestions for the next order. The entire system consists of a server, a terminal, and a user.

[0980] Hardware and software used

[0981] Speech Recognition: DeepSpeech, Google Speech-to-Text API

[0982] Generative AI model: GPT-4

[0983] Emotion recognition engine: IBM Watson Tone Analyzer

[0984] Cloud storage: Amazon S3

[0985] Database: MySQL, Firebase

[0986] Web interface: React.js, Node.js

[0987] Processing Details

[0988] 1. Review upload function

[0989] After ordering from a food delivery service, users record their impressions in audio format and upload them to the app. For example, a user might upload a comment such as, "The food I ordered was delicious, but the delivery was slow."

[0990] 2. Review Analysis and Emotion Recognition

[0991] The server receives the audio files uploaded by users and stores them in Amazon S3. It then converts the audio data into text using DeepSpeech or the Google Speech-to-Text API and stores it in a MySQL database. The converted text data undergoes sentiment analysis using IBM Watson Tone Analyzer to identify the user's emotional data.

[0992] 3. Proposal generation for next order

[0993] The next time the user logs in to the app and attempts to place an order, the server generates a prompt based on the stored emotional and text data. For example, "Based on the user's review, the last order was delayed but the user was satisfied with the taste. Generate a message with highlighted menu items and a discount for the next order based on this information." This prompt is input into GPT-4, which generates an appropriate message and suggestions. The generated message and suggestions are then displayed on the user's order screen.

[0994] Specific examples

[0995] For example, if a user leaves a review about their last order saying, "The food was delicious, but the delivery was late and I was a little disappointed," the next time they order, they might see a suggestion like this: "We apologize for the delay with your last order. To ensure we can deliver your order on time, please take a look at our new Chef's Specials. Plus, we'll give you a 10% discount on your next order."

[0996] How the program is implemented

[0997] The server uses the above-mentioned methods to analyze voice data, generate emotion recognition data, generate prompts based on the stored data, and output messages and suggestions using a generative AI model. This system makes it possible to automatically generate appropriate responses and suggestions that are linked to the user's emotions.

[0998] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0999] Step 1:

[1000] A user uses a food delivery service and records an audio review after placing an order. The user opens the application, uses the recording function to record their review, and uploads the audio file (e.g., "foodie_review_2023_10_01.wav") to the app. The input is the user's voice data, which is then stored in cloud storage.

[1001] Step 2:

[1002] The server receives the audio file uploaded by the user and stores it in cloud storage (Amazon S3). Next, the server invokes a speech recognition engine (DeepSpeech or Google Speech-to-Text API) to convert the audio data into text data. In this case, the input is the audio file and the output is text data. The converted text data (e.g., "The food was delicious, but the delivery was slow") is stored in a temporary database (MySQL).

[1003] Step 3:

[1004] The server uses an emotion recognition engine (IBM Watson Tone Analyzer) to analyze the converted text data and recognize the user's emotional data. In this process, the input is text data, and the output is the user's emotional data (e.g., "dissatisfied" or "satisfied"). The analysis results are stored in cloud storage or a database.

[1005] Step 4:

[1006] The next time the user logs into the app and attempts to place a new order, the server retrieves past reviews and sentiment data. The server generates a prompt based on this data and sends it to a generative AI model (GPT-4) to generate an appropriate message or suggestion. The input is past text data and sentiment data, and the output is the generated message or suggestion. An example of a prompt is, "According to the user's review, the food was delayed on their last order, but they were satisfied with the taste. Based on this information, please generate a message including featured menu items and discounts for their next order."

[1007] Step 5:

[1008] The generative AI model (GPT-4) generates new suggestions and messages based on the prompt. For example, "We apologize for the delay in your last order. As a special favor, we invite you to try our new menu item. We'll give you a 10% discount on your next order." The input is the prompt, and the output is a personalized message.

[1009] Step 6:

[1010] The server displays the generated message on the user's order screen. The user can then place a new order based on the displayed message and suggestions. The input is the generated message, and the output is the user's display screen. This improves the user experience and increases ordering satisfaction.

[1011] 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.

[1012] 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.

[1013] 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.

[1014] [Third embodiment]

[1015] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[1016] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[1017] 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).

[1018] 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.

[1019] 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.

[1020] 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).

[1021] 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. 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.

[1022] 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.

[1023] 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.

[1024] 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.

[1025] 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.

[1026] 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."

[1027] The system of the present invention uses a generative AI model to create minutes of parliamentary sessions and automatically generate materials for the next session (such as draft answers) based on that data. Below, we will explain how this system is implemented in concrete terms.

[1028] Acquisition of parliamentary recording data

[1029] The user uploads a new session recording file to the server. For example, after a session ends, the user uploads the recording data (e.g., "session_2023_10.wav") via the web interface. The file is received by the server and stored in cloud storage.

[1030] Data analysis and minutes preparation

[1031] The server analyzes the uploaded recording file. A speech recognition engine is used for this analysis, converting the audio data into text data. Speech recognition technologies such as DeepSpeech and Google Speech-to-Text API are used. The converted text data is split into segments for each speaker and appropriately formatted using speaker identification technology. It is then converted into a format (e.g., JSON) that is easy for the generative AI model to understand. Here is an example:

[1032] json

[1033] {

[1034] "session_id": "2023_10",

[1035] "timestamp": "2023-10-15",

[1036] "content": [

[1037] {

[1038] "speaker": "Rep. A",

[1039] "statement": "I'd like to ask about this year's budget."

[1040] },

[1041] {

[1042] "speaker": "Mayor B",

[1043] "statement": "The reason for the budget cuts is..."

[1044] }

[1045] ]

[1046] }

[1047] Storage of past meeting minutes data

[1048] The server saves the formatted minutes data in cloud storage. When saving, metadata (e.g., session ID, date) is also saved to ensure association with past minutes data.

[1049] Generate next parliamentary response

[1050] The user inputs the topic and questions for the next assembly meeting on the terminal. For example, the user might input, "I plan to ask a question about the budget at the next assembly meeting." The server then searches past minutes data based on this input and extracts relevant information.

[1051] The server uses a generative AI model (e.g., GPT-4) to generate new draft answers based on the extracted information. The generated draft answers are displayed on the device for the user to review and edit. For example, the following draft answers may be generated:

[1052] json

[1053] {

[1054] "suggestions": [

[1055] {

[1056] "topic": "budget",

[1057] "answer": "The budget cuts made last year were made to improve efficiency and were implemented in a way that minimized the impact on the educational environment."

[1058] }

[1059] ]

[1060] }

[1061] Natural language search function

[1062] A user can search the database of past meeting minutes using a natural language search query. For example, they can type "Show me last year's education budget discussion." The server analyzes this natural language query and searches for relevant meeting minutes data. As a result, the following search results are displayed on the user's device:

[1063] text

[1064] Meeting on October 15, 2022:

[1065] Councilman A: "Please explain last year's education budget cuts."

[1066] Mayor B: "The reason behind the budget cuts is..."

[1067] Parliamentary recording and video recording equipment

[1068] Users can also use audio and video recording equipment specifically designed for parliamentary use. This equipment is designed to record high-quality audio and video recordings of parliamentary speeches and debates. The audio and video recordings are then uploaded to a server for later analysis and storage.

[1069] This system allows local government employees to carry out their work efficiently and significantly reduces the time it takes to prepare assembly response proposals. In addition, by utilizing past meeting minutes data, it is possible to prepare consistent response proposals.

[1070] The processing flow will be explained below.

[1071] Step 1:

[1072] The user uploads the parliamentary recording (e.g., "session_2023_10.wav") to the server using the interface. The user selects the recording using a web browser and clicks the "Upload" button.

[1073] Step 2:

[1074] The server receives the uploaded recording data. After receiving it, the file is stored in cloud storage, where the file's metadata (e.g., session ID, upload date and time) is also recorded.

[1075] Step 3:

[1076] The server analyzes the received recording data, activates a speech recognition engine, and converts the audio data into text data, which is then stored in a temporary database.

[1077] Step 4:

[1078] The server extracts the text data obtained from the speech recognition engine and uses speaker identification technology to segment it into segments for each speaker, identifying speakers based on, for example, voice patterns or time of day.

[1079] Step 5:

[1080] The server formats the text data of the identified speaker into a format that is easy for the AI ​​model to understand (e.g., JSON format). The formatted data is then stored in cloud storage. The following is an example of the data:

[1081] json

[1082] {

[1083] "session_id": "2023_10",

[1084] "timestamp": "2023-10-15",

[1085] "content": [

[1086] {

[1087] "speaker": "Rep. A",

[1088] "statement": "I'd like to ask about this year's budget."

[1089] },

[1090] {

[1091] "speaker": "Mayor B",

[1092] "statement": "The reason for the budget cuts is..."

[1093] }

[1094] ]

[1095] }

[1096] Step 6:

[1097] The user enters the topic and question for the next session of the Diet on the device. For example, they enter, "I plan to ask a question about the budget at the next session of the Diet." The input is done through a web interface.

[1098] Step 7:

[1099] The server searches for past minutes data based on the input topic, and efficiently extracts related minutes data using metadata.

[1100] Step 8:

[1101] The server uses a generative AI model (e.g., GPT-4) to generate a new draft response based on the extracted minutes data. Below is an example of a generated draft response:

[1102] json

[1103] {

[1104] "suggestions": [

[1105] {

[1106] "topic": "budget",

[1107] "answer": "The budget cuts made last year were made to improve efficiency and were implemented in a way that minimized the impact on the educational environment."

[1108] }

[1109] ]

[1110] }

[1111] Step 9:

[1112] The generated draft answer is displayed on the device, where the user can review it and edit it if necessary. The final draft answer can also be saved in a different format or printed.

[1113] Step 10:

[1114] A user enters a natural language search query, for example, "Show me last year's education budget discussions."

[1115] Step 11:

[1116] The server analyzes the entered search query and searches the database for relevant meeting minutes data, using natural language processing techniques.

[1117] Step 12:

[1118] The search results are displayed on the device. For example, the following search results are displayed on the user's screen:

[1119] text

[1120] Meeting on October 15, 2022:

[1121] Councilman A: "Please explain last year's education budget cuts."

[1122] Mayor B: "The reason behind the budget cuts is..."

[1123] Example 1

[1124] 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."

[1125] Conventional methods for managing parliamentary recording data require a great deal of time and effort for manual minutes creation and data search, resulting in low productivity. It is also difficult to quickly propose consistent content for the next parliamentary session. Therefore, there is a need for a system that can automatically and efficiently create minutes and provide draft answers using generative AI models.

[1126] 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.

[1127] In this invention, the server includes means for receiving recorded data, means for analyzing the received recorded data and converting the audio data into text data, means for formatting the converted text data into a predetermined format, means for saving the formatted text data, means for generating a draft answer using a generative AI model based on previously saved text data, means for outputting the generated draft answer, and means for displaying the output draft answer on a terminal for user input. This makes it possible to automatically and efficiently create minutes and quickly propose consistent draft answers using the generative AI model.

[1128] "Recorded data" refers to data that records in audio format statements and discussions at meetings such as parliaments.

[1129] "Means for receiving" refers to the function or device that allows the server to import the recorded data uploaded by the user.

[1130] "Means for analyzing" refers to the technology or method for analyzing the received recorded data and converting the audio data into text data.

[1131] "Audio data" refers to audio signals stored as recorded data.

[1132] "Text data" is data of character strings that are obtained by analyzing voice data and converting it into linguistic information.

[1133] The "predetermined format" refers to the form or structure used to format the generated text data to ensure consistency and ease of understanding.

[1134] "Storage means" refers to a function or system for storing and managing formatted text data.

[1135] A "generative AI model" refers to a model that uses artificial intelligence to generate responses or suggestions based on specific input data.

[1136] A "draft answer" is a prepared answer that a member of parliament or other relevant person can use to appropriately respond to a question asked in a parliament or other such setting.

[1137] "Means of output" refers to the functions and systems for presenting the generated text data and draft answers to the user.

[1138] A "user input terminal" is a device used by a user to enter data and check results.

[1139] A "natural language search query" refers to a question that a user enters in natural language to search past data.

[1140] "Query-based search means" refers to techniques or methods for retrieving relevant past data from an input natural language search query.

[1141] "Means for outputting search results" refers to the functions and systems for displaying search results to the user.

[1142] "Recording device" refers to a device or equipment for acquiring audio and video data.

[1143] This invention is a system that uses generative AI models to efficiently create parliamentary minutes and automatically generate materials for the next parliamentary session. The system starts with the user uploading audio data, then analyzes the audio data, converts it to text data, stores the data, searches it, and generates draft responses.

[1144] Acquisition of parliamentary recording data

[1145] A user uploads the recorded data to the server via a web interface. For example, after a session ends, the user selects the recorded data (e.g., "session_2023_10.wav") and clicks the upload button. The server receives the data as an HTTP POST request and saves it to cloud storage (e.g., Amazon S3). The server notifies the user when the file has been saved.

[1146] Data analysis and minutes preparation

[1147] The server passes the received recording data to a speech recognition engine, which converts the audio data into text data. This analysis is performed using, for example, the Google Speech-to-Text API. The audio data is converted into a byte stream and sent via the API. The returned text data is split into segments for each speaker using speaker identification technology and formatted into a JSON format that is easy for the generative AI model (e.g., GPT-4) to understand.

[1148] For example, the following JSON data is generated:

[1149] json

[1150] {

[1151] "session_id": "2023_10",

[1152] "timestamp": "2023-10-15",

[1153] "content": [

[1154] {

[1155] "speaker": "Rep. A",

[1156] "statement": "I'd like to ask about this year's budget."

[1157] },

[1158] {

[1159] "speaker": "Mayor B",

[1160] "statement": "The reason for the budget cuts is..."

[1161] }

[1162] ]

[1163] }

[1164] Storage of past meeting minutes data

[1165] The server saves the formatted JSON data in cloud storage. This allows it to be associated with past meeting minutes data. When saving, metadata (e.g., session ID, date) is also saved.

[1166] Generate next parliamentary response

[1167] The user uses an input form on their device to input the topic and questions for the next session. For example, they might input, "I plan to ask a question about the budget at the next session." The server receives this input and searches past minutes data to extract relevant information. It then uses a generative AI model (e.g., GPT-4) to generate new draft responses. An example of this prompt is, "Imagine a question about 'budget cuts' at the next session. Please generate a relevant draft response based on past minutes data." The generated draft response is displayed on the user's device.

[1168] Natural language search function

[1169] Users can search past meeting minutes data by entering natural language search queries into their devices. For example, they can enter "Show me last year's education budget discussions." The server analyzes this input using a natural language processing (NLP) module and searches for relevant data. Search results are displayed on the user's device.

[1170] You'll see specific search results like this:

[1171] text

[1172] Meeting on October 15, 2022:

[1173] Councilman A: "Please explain last year's education budget cuts."

[1174] Mayor B: "The reason behind the budget cuts is..."

[1175] Parliamentary recording and video recording equipment

[1176] Users operate dedicated audio and video recording equipment for council meetings to record speeches and discussions in high quality. The recorded data is later uploaded to a server for analysis and storage. This system enables local government officials to carry out their work efficiently and prepare consistent responses.

[1177] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1178] Step 1: Upload your recordings

[1179] The user uploads the recording (e.g., "session_2023_10.wav") to the server via a web interface. Input: The recording file of the parliament. Output: The uploaded recording file is saved in cloud storage. Specific behavior: The user opens a file selection dialog in a web browser, selects the recording data from the local file system, and presses the upload button. The server receives the HTTP POST request and saves the file in cloud storage.

[1180] Step 2: Analyzing the audio data

[1181] The server retrieves the recorded data from cloud storage and sends it to a speech recognition engine (e.g., Google Speech-to-Text API). Input: The saved recording file. Output: The speech data converted to text data. Specific operation: The server converts the speech data into a byte stream, sends it to the recognition engine, and receives the returned text data.

[1182] Step 3: Formatting the text data

[1183] The server analyzes the acquired text data and uses speaker identification technology to split and format it for each speaker. Input: Text data returned from the speech recognition engine. Output: Meeting minutes data in JSON format. Specific operation: The server analyzes the text data, splits it for each speaker, and generates formatted JSON data.

[1184] Step 4: Save your data

[1185] The server saves the formatted JSON data in cloud storage. Input: Formatted text data (JSON format). Output: Meeting minutes data saved in cloud storage. Specific operation: The server saves the JSON data in the specified directory, along with the metadata.

[1186] Step 5: Generate draft answers for the next session

[1187] The user inputs the topic and question for the next session of the Diet (e.g., "I plan to ask a question about the budget at the next session of the Diet") into the device. Input: The topic and question for the next session of the Diet. Output: Generated draft answers. Specific operation: The server searches past minutes data based on the input topic and question, generates new draft answers using a generative AI model (e.g., GPT-4), and displays them on the device.

[1188] Step 6: Natural Language Search

[1189] A user enters a natural language search query (e.g., "Show me last year's education budget discussions") into a device. Input: Natural language search query. Output: Relevant meeting minutes data. Specific behavior: The server uses a natural language processing (NLP) module to parse the search query, search for relevant data, and display the search results on the device.

[1190] Step 7: Operating the recording equipment

[1191] The user operates audio and video recording equipment specifically designed for the assembly to capture high-quality audio and video data. Input: Data captured by the audio and video recording equipment. Output: The audio and video data is uploaded to the server and stored in cloud storage. Specific operation: The user presses the start recording button on the audio and video recording equipment, which records the entire assembly meeting and then uploads the data to the server.

[1192] (Application example 1)

[1193] 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."

[1194] In logistics centers, creating minutes of meetings and conferences is a time-consuming and labor-intensive process, and labor shortages can lead to inefficiencies. It is also difficult to quickly retrieve information related to the next meeting or task by searching past meeting records. The present invention aims to solve these problems by providing a system that automatically converts meeting recording data into text and generates materials and plans for the next meeting based on a generative AI model.

[1195] 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.

[1196] In this invention, the server includes means for receiving recorded assembly data, means for analyzing the received recorded data and converting the audio data into text data, means for formatting the converted text data into a predetermined format, means for saving the formatted text data, means for generating draft responses using a generative AI model based on previously saved text data, means for inputting the next meeting theme and tasks, means for extracting related information and generating new meeting materials and draft plans, and means for outputting the generated draft responses and meeting materials. This automates the creation of meeting minutes within the logistics center, enabling the efficiency of preparations for the next meeting and tasks.

[1197] "Parliamentary recording data" refers to data files containing recordings of speeches and discussions held in parliaments or meetings.

[1198] "Means for receiving" refers to the function for obtaining recorded data from outside and importing it into the server.

[1199] "Means for analyzing and converting voice data into text data" refers to a function for converting recorded voice data into text information using voice recognition technology.

[1200] "Means for formatting into a specified format" refers to a function for arranging converted text data according to certain rules and formats.

[1201] "Means of storage" refers to the function of storing formatted text data in a database, cloud storage, etc. for future reference or search.

[1202] "Means for generating draft responses using generative AI models" refers to a function that uses AI technology based on past data to automatically create draft responses to be used in the next parliament or meeting.

[1203] "Means for inputting the next meeting topic and tasks" refers to an interface that allows the user to provide the server with information and questions about the next meeting.

[1204] "Means for extracting relevant information and generating new materials and plans" refers to the function of extracting necessary information from past data and using it to create new materials and plans.

[1205] "Means for outputting" refers to a function for providing the generated draft answers and meeting materials to the user.

[1206] "Means for accepting natural language search queries" refers to an interface that allows users to enter search terms in everyday language.

[1207] "Means for searching previously stored text data based on a search query" refers to a function for extracting data that matches the user's input from a database.

[1208] "Means for outputting search results" refers to a function for providing the searched content to the user.

[1209] "Means for additionally analyzing recorded data" refers to a function for further analyzing recorded data within a logistics facility and converting it into text data, etc.

[1210] "Function to record meeting recording data" refers to the function to collect the contents of meetings and assemblies in audio format.

[1211] "The function of recording recorded data" refers to the function of collecting the contents of meetings and assemblies in video format.

[1212] "Recording and video recording equipment" refers to devices and equipment used to record the contents of meetings and assemblies as audio or video.

[1213] MODE FOR CARRYING OUT THE INVENTION

[1214] This system automatically analyzes recorded meeting data at a logistics center, converts it into text data, and saves it. It can also automatically generate materials and plans for the next meeting using a generative AI model. It also has the ability to search past meeting records using natural language search queries and provide the results to users.

[1215] System Overview

[1216] The system consists of the following main components:

[1217] 1. Means of receiving parliamentary recordings

[1218] 2. Means of converting audio data into text data

[1219] 3. A means of formatting text data into a specified format

[1220] 4. Means of saving text data

[1221] 5. How to generate draft responses using generative AI models

[1222] 6. A way to enter the next meeting topic or task

[1223] 7. A means to extract relevant information and generate new meeting materials and plans

[1224] 8. Means for outputting the generated response plan and meeting materials

[1225] 9. A way to accept natural language search queries

[1226] 10. A way to search past data based on a search query

[1227] 11. How to output search results

[1228] 12. Additional means of analyzing recording data

[1229] 13. Function to record conference recording data

[1230] 14. Audio and video recording devices with the function of recording video data

[1231] System Operation

[1232] The server receives the parliamentary recording data, analyzes it, and converts the audio data into text data. This speech recognition uses technologies such as DeepSpeech and Google Speech-to-Text API. The converted text data is formatted and stored in cloud storage (e.g., AWS S3).

[1233] Users can input the topic or task for the next meeting into their device. For example, they can input a prompt such as, "Please create a document to discuss the logistics situation at the next meeting." The server then searches past text data and extracts relevant information. Based on this extracted information, a generative AI model (e.g., GPT-4) is used to generate new meeting materials and draft plans. The generated materials and draft responses are then displayed on the user's device.

[1234] Additionally, users can search for past meeting records using natural language search queries, such as a prompt such as "Show me meeting records from last year regarding logistics status." The server analyzes the query, searches for relevant data, and provides the results to the user.

[1235] Hardware and software used

[1236] The system uses the following hardware and software:

[1237] Server: Amazon Web Services (AWS) EC2

[1238] Cloud storage: AWS S3

[1239] Speech recognition engine: DeepSpeech, Google Speech-to-Text API

[1240] Generative AI model: OpenAI GPT-4

[1241] Natural language processing libraries: SpaCy, NLTK

[1242] Specific examples

[1243] For example, suppose a logistics center manager holds a weekly meeting and records it. The recording is then uploaded to a server, which analyzes the data and creates meeting minutes. For next week's meeting, if a user types into their device, "Please generate this week's plan based on last week's minutes," the server will search past data and automatically generate a new plan using a generative AI model. This can improve work efficiency.

[1244] Examples of prompts include "Please prepare materials to talk about the logistics situation at the next meeting" and "Show me the minutes of last year's meeting about the logistics situation."

[1245] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1246] Step 1:

[1247] A user acquires parliamentary recording data using audio and video recording equipment and uploads the data to the server. In this example, the recording file (e.g., "meeting_2023_10.wav") is sent to the server via a web interface. The input is an audio file, and the output is an audio file stored in cloud storage.

[1248] Step 2:

[1249] The server analyzes the received recording data and converts the audio data into text data. This process uses a speech recognition engine (DeepSpeech or Google Speech-to-Text API). The input is an audio file, and the output is the converted text data.

[1250] Step 3:

[1251] The server formats the converted text data into a specified format. Specifically, it uses speaker identification technology to split the text data for each speaker and converts it into JSON format. The input is raw text data, and the output is formatted text data.

[1252] Step 4:

[1253] The server saves the formatted text data in cloud storage (AWS S3), along with metadata (e.g., session ID, date). The input is the formatted text data, and the output is the data saved in cloud storage.

[1254] Step 5:

[1255] The user inputs the next meeting topic and task into the terminal. For example, the user might input, "Please prepare materials to talk about the logistics situation at the next meeting." The input is a prompt text, and the output is sending the prompt text to the server.

[1256] Step 6:

[1257] The server searches for information related to the prompt text entered by the user based on previously saved text data. The input is the prompt text and past text data, and the output is the searched related information.

[1258] Step 7:

[1259] Based on the related information retrieved by the server, a generative AI model (GPT-4) is used to generate meeting materials and plan proposals. The input is related information, and the output is new meeting materials and plan proposals.

[1260] Step 8:

[1261] The server outputs the generated meeting materials and plans to the terminals to provide them to the users. The input is the new meeting materials and plans, and the output is the materials and plans displayed on the user's terminal.

[1262] Step 9:

[1263] A user inputs a search query in natural language into a terminal. For example, "Show me the minutes of the meeting about the logistics situation last year." The input is a search query in natural language, and the output is a search query sent to the server.

[1264] Step 10:

[1265] The server searches past text data based on the search query. Here, it references a database of past meeting minutes and extracts relevant data. The input is the search query and past text data, and the output is the search result data.

[1266] Step 11:

[1267] The server outputs the search results to the terminal, allowing the user to view the results. The input is the search result data, and the output is the search result displayed on the user's terminal.

[1268] 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.

[1269] The system of the present invention uses a generative AI model to create minutes of parliamentary sessions and automatically generates the next parliamentary session materials (such as draft parliamentary responses) based on that data. Furthermore, by combining it with an emotion engine that recognizes the user's emotions, it is possible to generate more appropriate draft responses that reflect the user's intentions and emotions. Below, we will explain in detail how this system is implemented.

[1270] Acquisition of parliamentary recording data

[1271] The user uploads the parliamentary recording (e.g., "session_2023_10.wav") to the server using the interface. The user selects the recording via a web browser and clicks the "Upload" button. The file is received by the server and stored in cloud storage.

[1272] Data analysis and minutes preparation

[1273] The server analyzes the uploaded recording data, activates a speech recognition engine, and converts the audio data into text data. The speech recognition technology used is DeepSpeech or the Google Speech-to-Text API, for example. The converted text data is then stored in a temporary database.

[1274] Speaker Identification and Data Shaping

[1275] The server extracts the text data obtained from the speech recognition engine and divides it into segments for each speaker using speaker identification technology. For example, speakers can be identified based on voice patterns or time of day. The server then formats the text data of the identified speaker into a format (e.g., JSON) that is easy for the generative AI model to understand. The formatted data is then saved in cloud storage. The following is an example of the data:

[1276] json

[1277] {

[1278] "session_id": "2023_10",

[1279] "timestamp": "2023-10-15",

[1280] "content": [

[1281] {

[1282] "speaker": "Rep. A",

[1283] "statement": "I'd like to ask about this year's budget."

[1284] },

[1285] {

[1286] "speaker": "Mayor B",

[1287] "statement": "The reason for the budget cuts is..."

[1288] }

[1289] ]

[1290] }

[1291] Generating Answer Plans and Using an Emotion Engine

[1292] The user inputs the topic and questions for the next session of the Diet on the terminal. For example, the user might input, "I plan to ask a question about the budget at the next session of the Diet." The input is done via a web interface. The server searches past minutes data based on this input and extracts relevant information.

[1293] The server uses a generative AI model (e.g., GPT-4) to generate new response proposals based on the extracted information. Furthermore, an emotion engine recognizes the user's emotional data and adjusts the response proposals based on that data. For example, if the user is recognized as "nervous," the response proposals will include expressions that convey a more reassuring feeling.

[1294] json

[1295] {

[1296] "suggestions": [

[1297] {

[1298] "topic": "budget",

[1299] "answer": "The budget cuts made last year were made to improve efficiency and we have taken measures to minimize the impact on the educational environment. Please rest assured."

[1300] }

[1301] ]

[1302] }

[1303] Storing and using emotional data

[1304] The server stores the user's emotional data recognized by the emotion engine in cloud storage. The stored emotional data is used to generate future parliamentary response drafts. This results in more consistent response drafts and increases user confidence.

[1305] Natural language search function

[1306] A user can search the database of past meeting minutes using a natural language search query. For example, they can type "Show me last year's education budget discussion." The server analyzes this natural language query and searches for relevant meeting minutes data. As a result, the following search results are displayed on the user's device:

[1307] text

[1308] Meeting on October 15, 2022:

[1309] Councilman A: "Please explain last year's education budget cuts."

[1310] Mayor B: "The reason behind the budget cuts is..."

[1311] Parliamentary recording and video recording equipment

[1312] Users can record high-quality audio and video of parliamentary speeches and debates using dedicated parliamentary audio and video recording equipment, which will then be uploaded to a server for later analysis and storage.

[1313] This system allows local government employees to carry out their work efficiently and significantly reduces the time it takes to prepare assembly response drafts. Furthermore, by taking into consideration the user's feelings, it is possible to provide more reliable response drafts. In addition, by utilizing past minutes data, it is possible to prepare consistent response drafts.

[1314] The processing flow will be explained below.

[1315] Step 1:

[1316] The user uploads the parliamentary recording (e.g., "session_2023_10.wav") to the server using the interface. The user selects the recording via a web browser and clicks the "Upload" button.

[1317] Step 2:

[1318] The server receives the uploaded recording data. After receiving it, the file is stored in cloud storage, where the file's metadata (e.g., session ID, upload date and time) is also recorded.

[1319] Step 3:

[1320] The server analyzes the received recording, invokes a speech recognition engine (e.g., DeepSpeech or Google Speech-to-Text API), and converts the audio data into text data, which is then stored in a temporary database.

[1321] Step 4:

[1322] The server extracts the text data obtained from the speech recognition engine and uses speaker identification technology to segment it into segments for each speaker, identifying speakers based on, for example, voice patterns or time of day.

[1323] Step 5:

[1324] The server formats the text data of the identified speaker into a format that is easy for the AI ​​model to understand (e.g., JSON format). The formatted data is then stored in cloud storage. The following is an example of the data:

[1325] json

[1326] {

[1327] "session_id": "2023_10",

[1328] "timestamp": "2023-10-15",

[1329] "content": [

[1330] {

[1331] "speaker": "Rep. A",

[1332] "statement": "I'd like to ask about this year's budget."

[1333] },

[1334] {

[1335] "speaker": "Mayor B",

[1336] "statement": "The reason for the budget cuts is..."

[1337] }

[1338] ]

[1339] }

[1340] Step 6:

[1341] The user inputs the topic and questions for the next council meeting on the terminal. For example, the user inputs, "I plan to ask a question about the budget at the next council meeting." The input is done via a web interface.

[1342] Step 7:

[1343] The server searches for past minutes data based on the input topic, and efficiently extracts related minutes data using metadata.

[1344] Step 8:

[1345] The server uses a generative AI model (e.g., GPT-4) to generate a new draft response based on the extracted minutes data. Below is an example of a generated draft response:

[1346] json

[1347] {

[1348] "suggestions": [

[1349] {

[1350] "topic": "budget",

[1351] "answer": "The budget cuts made last year were made to improve efficiency and were implemented in a way that minimized the impact on the educational environment."

[1352] }

[1353] ]

[1354] }

[1355] Step 9:

[1356] The server uses the emotion engine to recognize the user's emotion, for example, the emotion engine recognizes that the user is "nervous."

[1357] Step 10:

[1358] The server reflects the emotional data it recognizes in the generated answer proposals. For example, if the server recognizes that the user is "nervous," the answer proposals include expressions that make the user feel more at ease.

[1359] Step 11:

[1360] The generated draft answer is displayed on the device, where the user can review it and edit it if necessary. The final draft answer can also be saved in a different format or printed.

[1361] Step 12:

[1362] A user enters a natural language search query, for example, "Show me last year's education budget discussions."

[1363] Step 13:

[1364] The server analyzes the entered search query and searches the database for relevant meeting minutes data, using natural language processing techniques.

[1365] Step 14:

[1366] The search results are displayed on the device. For example, the following search results are displayed on the user's screen:

[1367] text

[1368] Meeting on October 15, 2022:

[1369] Councilman A: "Please explain last year's education budget cuts."

[1370] Mayor B: "The reason behind the budget cuts is..."

[1371] Example 2

[1372] 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."

[1373] The task of creating minutes of assembly meetings and response materials for the next assembly meeting requires time and effort, placing a heavy burden on staff. Furthermore, creating response materials requires appropriately reflecting the content of past discussions while also taking into account the user's emotions and intentions, but there is no efficient way to do this. In conventional technologies, creating minutes and generating response materials is often done manually, resulting in reduced work efficiency and inconsistent quality. This invention is intended to solve these problems.

[1374] The specific processing by the specific processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for receiving parliamentary recording data, means for analyzing the received recording data and converting the audio data into text data, means for formatting the converted text data into a predetermined data format, means for generating a draft response using a generative AI model based on previously saved text data, means for reflecting the user's emotions in the generated draft response using an emotion engine, and means for outputting the generated draft response. This automates the creation of minutes and the generation of response materials, improving work efficiency and ensuring consistent quality. Furthermore, because response materials are generated taking the user's emotions into consideration, more appropriate and reliable materials can be created.

[1375] "Parliamentary recording data" refers to audio data of speeches and discussions in parliament.

[1376] A "generative AI model" refers to an artificial intelligence model that generates new outputs based on input data.

[1377] "Emotion engine" refers to a program or function that recognizes a user's emotions and analyzes the emotion data.

[1378] "Draft answers" refers to prepared answers to questions asked in parliament.

[1379] "Converted text data" refers to character data converted from voice data using voice recognition technology.

[1380] "Data format" refers to the format or standard for organizing and arranging data in a certain structure.

[1381] A "natural language search query" refers to a search request made to a database using natural language.

[1382] "Sound recording equipment" means devices for recording sound and video.

[1383] "Cloud storage" refers to an online storage service that stores and manages data via the Internet.

[1384] A "server" refers to a computer system that provides services and data to other computers over a network.

[1385] "Terminal" refers to a device used by a user for input and display.

[1386] "Text data" refers to data expressed as character information.

[1387] The system of the present invention utilizes a generative AI model and an emotion engine to create minutes of parliamentary meetings and automatically generate materials for responses to questions at the next parliamentary meeting. A detailed embodiment of this system will be described.

[1388] Overall system overview

[1389] This system analyzes recordings of parliamentary sessions, converts the audio data into text data, and uses that text data to automatically generate draft responses for the next session. It can also recognize the user's emotions and generate draft responses that reflect those emotions. This system consists of three main components: a server, a terminal, and a user.

[1390] Hardware and Software Configuration

[1391] The server receives the recorded data and analyzes the audio data. It is equipped with a high-performance processor and a large amount of memory. The speech recognition engine uses common speech recognition software such as DeepSpeech or Google Speech-to-Text API. The server also converts the analyzed text data into a specified data format and stores it in a temporary database.

[1392] The terminal provides an interface for users to input the topic and questions for the next session. This terminal is a standard PC or tablet and is connected to the server via a web browser. Users can use the terminal to upload recording data as well as check the generated response plans.

[1393] Users upload the recorded data of assembly meetings to the server using their devices. They input the topic and questions of the next assembly meeting, and check and use the proposed answers provided by the server. Users can also input search queries to search for past minutes data.

[1394] Processing flow and specific examples

[1395] The system begins by a user uploading a recording of a parliamentary session. The recording (e.g., "session_2023_10.wav") is then stored in cloud storage by the server. The server then activates a speech recognition engine, analyzes the recording, and converts the audio data into text data. This text data is then split by speaker using speaker identification technology and formatted into a format that is easy for the generative AI model to understand.

[1396] For example, a statement such as "I'd like to ask about this year's budget" is linked to "Council Member A," and a statement such as "The reason for the budget cuts is..." is linked to "Mayor B." This results in the data being formatted as text like this:

[1397] Session ID: 2023_10

[1398] Timestamp: 2023-10-15

[1399] Contents:

[1400] Councilman A: "I'd like to ask about this year's budget."

[1401] Mayor B: 'The reason for the budget cuts is...'

[1402] "

[1403] Next, the user inputs the next session's theme and question on the device. For example, they might input, "I plan to ask a question about the budget at the next session." The server extracts relevant information from past minutes data and uses this information to input prompts into the generative AI model to generate new draft responses. The emotion engine also recognizes the user's emotions, and if it recognizes that the user is "nervous," the generated draft responses will include expressions that convey a sense of security.

[1404] Examples of prompts are:

[1405] "Parliamentary Data:

[1406] Session ID: 2023_10

[1407] Timestamp: 2023-10-15

[1408] Contents:

[1409] Councilman A: "I'd like to ask about this year's budget."

[1410] Mayor B: 'The reason for the budget cuts is...'

[1411] Next Congress Theme and User Sentiments:

[1412] Theme: 'Budget'

[1413] Question: I plan to ask a question about the budget at the next Congress. I would like to know specifically why the budget cuts were made.

[1414] Emotion: 'Nervous'

[1415] Generated answer plan:

[1416] Topic: 'Budget'

[1417] Response: 'The budget cuts made last year were made to improve efficiency and we have taken measures to minimize the impact on the educational environment. Please rest assured.'

[1418] "

[1419] The server saves the generated answer plan in cloud storage and displays it on the user's device. The user's emotional data is also saved in cloud storage and used to generate answer plans from the next time onwards, making it possible to create consistent answer plans.

[1420] The above is a specific embodiment of the system. This system automates the creation of assembly minutes and the generation of response materials for the next assembly meeting, and can provide high-quality materials that take user emotions into consideration.

[1421] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1422] Step 1: Obtaining Congressional Recording Data

[1423] Input: The user uploads a recording file (e.g., "session_2023_10.wav") through the designated interface.

[1424] How it works: A user uses a web browser to select a recording and presses the "upload" button.

[1425] Output: The server saves this recording to cloud storage and records the file's metadata (e.g., file name, upload date and time).

[1426] Step 2: Analyze the data and create a transcript

[1427] Input: Recording data files stored in cloud storage.

[1428] How it works: The server launches a speech recognition engine (e.g., DeepSpeech or Google Speech-to-Text API) and reads the recording. The server converts the audio data into text. This conversion process converts what is being said into text format.

[1429] Output: The converted text data is saved in a temporary database.

[1430] Step 3: Speaker identification and data conditioning

[1431] Input: Text data stored in a temporary database.

[1432] How it works: The server uses speaker identification technology to identify each speaker, using characteristics such as voice patterns and time of day. It then organizes the text data for each identified speaker and formats it into a data format (e.g., JSON) that is easy for the generative AI model to understand.

[1433] Output: The formatted text data is saved to cloud storage.

[1434] Step 4: Enter the theme and questions for the next assembly meeting

[1435] Input: The user enters the topic or question for the next session through a web interface (e.g., "I have a question about the budget for the next session").

[1436] How it works: The user uses a terminal to enter the required information and send it to the server.

[1437] Output: The server receives this input data and prepares it for the next processing step.

[1438] Step 5: Generate a response and use the emotion engine

[1439] Input: User questions and past meeting minutes data.

[1440] How it works: The server extracts relevant information from past meeting minutes data and inputs prompt sentences into a generative AI model. A generative AI model (e.g., GPT-4) is used to generate new response proposals. The emotion engine then analyzes the user's emotion data and adjusts the response proposals based on the recognized emotion. For example, "If the user is nervous, add expressions that will make them feel at ease."

[1441] Output: A response plan reflecting the sentiment is generated and saved in cloud storage.

[1442] Step 6: Outputting the generated answer plan

[1443] Input: Response plan data stored on the server.

[1444] Operation: The server sends the generated answer plan to the user's terminal, where the user can view the answer plan.

[1445] Output: The proposed answer displayed on the user's device.

[1446] Step 7: Storing and using emotion data

[1447] Input: User emotion data analyzed by the emotion engine.

[1448] How it works: The server stores the emotional data in cloud storage. This data is used to understand the user's emotional tendencies for future response generation.

[1449] Output: Emotion data stored in cloud storage.

[1450] Step 8: Processing Natural Language Search Queries

[1451] Input: A natural language search query entered by a user through a web interface (e.g., "Show me last year's education budget discussions").

[1452] How it works: The server analyzes the search query and searches for relevant information from past meeting minutes data.

[1453] Output: Search results displayed on the user's device.

[1454] In this way, each processing step works together to automate the entire process, from creating minutes of parliamentary sessions to generating response materials for the next session of parliament.

[1455] (Application example 2)

[1456] 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."

[1457] Conventional meeting minutes creation systems simply convert the minutes into text and store them, but there is a problem in that it takes a lot of time and effort to properly prepare for the next meeting and generate draft responses. It is also difficult to create draft responses that fully reflect the user's emotions and intentions, which can lead to a decrease in user satisfaction. Furthermore, there is a demand for similar technology to be applied to other industries, and in the food delivery field in particular, there is a strong need for technology that combines review analysis and emotion recognition to improve customer satisfaction.

[1458] 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.

[1459] In this invention, the server includes means for receiving parliamentary recording data, means for analyzing the received recording data and converting the audio data into text data, means for formatting the converted text data into a predetermined format, means for saving the formatted text data, means for generating draft responses using a generative AI model based on previously saved text data, means for analyzing a user's emotional data using an emotion recognition engine and adjusting the draft responses using the generative AI model based on this, means for generating personalized messages and suggestions that reflect the user's emotional data at the time of the next order, and means for outputting the generated draft responses. This makes it possible to automatically generate appropriate draft responses and suggestions that are linked to the user's emotions.

[1460] "Parliamentary recording data" refers to audio recordings of speeches and discussions in parliament.

[1461] "Means" refers to a method or apparatus for performing a specific function or process in a patent.

[1462] "Audio data" refers to digitalized sound information acquired through a voice input device such as a microphone.

[1463] "Text data" is a form of data expressed as character information.

[1464] A "format" is a set of rules for arranging data into a particular structure or form.

[1465] A "generative AI model" is an algorithm or program that uses machine learning technology to create new information.

[1466] A "draft reply" is a proposed response to a specific question or topic in parliament or elsewhere.

[1467] An "emotion recognition engine" is a technology that analyzes and identifies human emotions from data such as voice and text.

[1468] "User" refers to a person or organization that uses the system.

[1469] A "message" is a sentence or text intended to convey specific information.

[1470] A "suggestion" is an idea or opinion submitted recommending a particular action or option.

[1471] A "system" is a collection of integrated structures or devices in which multiple elements work together to perform a specific task.

[1472] System Overview

[1473] This system applies technology for analyzing recorded conference data to generate new suggestions and messages based on users' emotional data in order to improve customer satisfaction in food delivery services. The system allows users to upload audio reviews, analyzes the audio data to recognize emotions, and provides personalized messages and suggestions for the next order. The entire system consists of a server, a terminal, and a user.

[1474] Hardware and software used

[1475] Speech Recognition: DeepSpeech, Google Speech-to-Text API

[1476] Generative AI model: GPT-4

[1477] Emotion recognition engine: IBM Watson Tone Analyzer

[1478] Cloud storage: Amazon S3

[1479] Database: MySQL, Firebase

[1480] Web interface: React.js, Node.js

[1481] Processing Details

[1482] 1. Review upload function

[1483] After ordering from a food delivery service, users record their impressions in audio format and upload them to the app. For example, a user might upload a comment such as, "The food I ordered was delicious, but the delivery was slow."

[1484] 2. Review Analysis and Emotion Recognition

[1485] The server receives the audio files uploaded by users and stores them in Amazon S3. It then converts the audio data into text using DeepSpeech or the Google Speech-to-Text API and stores it in a MySQL database. The converted text data undergoes sentiment analysis using IBM Watson Tone Analyzer to identify the user's emotional data.

[1486] 3. Proposal generation for next order

[1487] The next time the user logs in to the app and attempts to place an order, the server generates a prompt based on the stored emotional and text data. For example, "Based on the user's review, the last order was delayed but the user was satisfied with the taste. Generate a message with highlighted menu items and a discount for the next order based on this information." This prompt is input into GPT-4, which generates an appropriate message and suggestions. The generated message and suggestions are then displayed on the user's order screen.

[1488] Specific examples

[1489] For example, if a user leaves a review about their last order saying, "The food was delicious, but the delivery was late and I was a little disappointed," the next time they order, they might see a suggestion like this: "We apologize for the delay with your last order. To ensure we can deliver your order on time, please take a look at our new Chef's Specials. Plus, we'll give you a 10% discount on your next order."

[1490] How the program is implemented

[1491] The server uses the above-mentioned methods to analyze voice data, generate emotion recognition data, generate prompts based on the stored data, and output messages and suggestions using a generative AI model. This system makes it possible to automatically generate appropriate responses and suggestions that are linked to the user's emotions.

[1492] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1493] Step 1:

[1494] A user uses a food delivery service and records an audio review after placing an order. The user opens the application, uses the recording function to record their review, and uploads the audio file (e.g., "foodie_review_2023_10_01.wav") to the app. The input is the user's voice data, which is then stored in cloud storage.

[1495] Step 2:

[1496] The server receives the audio file uploaded by the user and stores it in cloud storage (Amazon S3). Next, the server invokes a speech recognition engine (DeepSpeech or Google Speech-to-Text API) to convert the audio data into text data. In this case, the input is the audio file and the output is text data. The converted text data (e.g., "The food was delicious, but the delivery was slow") is stored in a temporary database (MySQL).

[1497] Step 3:

[1498] The server uses an emotion recognition engine (IBM Watson Tone Analyzer) to analyze the converted text data and recognize the user's emotional data. In this process, the input is text data, and the output is the user's emotional data (e.g., "dissatisfied" or "satisfied"). The analysis results are stored in cloud storage or a database.

[1499] Step 4:

[1500] The next time the user logs into the app and attempts to place a new order, the server retrieves past reviews and sentiment data. The server generates a prompt based on this data and sends it to a generative AI model (GPT-4) to generate an appropriate message or suggestion. The input is past text data and sentiment data, and the output is the generated message or suggestion. An example of a prompt is, "According to the user's review, the food was delayed on their last order, but they were satisfied with the taste. Based on this information, please generate a message including featured menu items and discounts for their next order."

[1501] Step 5:

[1502] The generative AI model (GPT-4) generates new suggestions and messages based on the prompt. For example, "We apologize for the delay in your last order. As a special favor, we invite you to try our new menu item. We'll give you a 10% discount on your next order." The input is the prompt, and the output is a personalized message.

[1503] Step 6:

[1504] The server displays the generated message on the user's order screen. The user can then place a new order based on the displayed message and suggestions. The input is the generated message, and the output is the user's display screen. This improves the user experience and increases ordering satisfaction.

[1505] 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.

[1506] 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.

[1507] 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.

[1508] [Fourth embodiment]

[1509] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1510] 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.

[1511] 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).

[1512] 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.

[1513] 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.

[1514] 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).

[1515] 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. 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.

[1516] 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.

[1517] 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.

[1518] 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.

[1519] 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.

[1520] 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.

[1521] 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."

[1522] The system of the present invention uses a generative AI model to create minutes of parliamentary sessions and automatically generate materials for the next session (such as draft answers) based on that data. Below, we will explain how this system is implemented in concrete terms.

[1523] Acquisition of parliamentary recording data

[1524] The user uploads a new session recording file to the server. For example, after a session ends, the user uploads the recording data (e.g., "session_2023_10.wav") via the web interface. The file is received by the server and stored in cloud storage.

[1525] Data analysis and minutes preparation

[1526] The server analyzes the uploaded recording file. A speech recognition engine is used for this analysis, converting the audio data into text data. Speech recognition technologies such as DeepSpeech and Google Speech-to-Text API are used. The converted text data is split into segments for each speaker and appropriately formatted using speaker identification technology. It is then converted into a format (e.g., JSON) that is easy for the generative AI model to understand. Here is an example:

[1527] json

[1528] {

[1529] "session_id": "2023_10",

[1530] "timestamp": "2023-10-15",

[1531] "content": [

[1532] {

[1533] "speaker": "Rep. A",

[1534] "statement": "I'd like to ask about this year's budget."

[1535] },

[1536] {

[1537] "speaker": "Mayor B",

[1538] "statement": "The reason for the budget cuts is..."

[1539] }

[1540] ]

[1541] }

[1542] Storage of past meeting minutes data

[1543] The server saves the formatted minutes data in cloud storage. When saving, metadata (e.g., session ID, date) is also saved to ensure association with past minutes data.

[1544] Generate next parliamentary response

[1545] The user inputs the topic and questions for the next assembly meeting on the terminal. For example, the user might input, "I plan to ask a question about the budget at the next assembly meeting." The server then searches past minutes data based on this input and extracts relevant information.

[1546] The server uses a generative AI model (e.g., GPT-4) to generate new draft answers based on the extracted information. The generated draft answers are displayed on the device for the user to review and edit. For example, the following draft answers may be generated:

[1547] json

[1548] {

[1549] "suggestions": [

[1550] {

[1551] "topic": "budget",

[1552] "answer": "The budget cuts made last year were made to improve efficiency and were implemented in a way that minimized the impact on the educational environment."

[1553] }

[1554] ]

[1555] }

[1556] Natural language search function

[1557] A user can search the database of past meeting minutes using a natural language search query. For example, they can type "Show me last year's education budget discussion." The server analyzes this natural language query and searches for relevant meeting minutes data. As a result, the following search results are displayed on the user's device:

[1558] text

[1559] Meeting on October 15, 2022:

[1560] Councilman A: "Please explain last year's education budget cuts."

[1561] Mayor B: "The reason behind the budget cuts is..."

[1562] Parliamentary recording and video recording equipment

[1563] Users can also use audio and video recording equipment specifically designed for parliamentary use. This equipment is designed to record high-quality audio and video recordings of parliamentary speeches and debates. The audio and video recordings are then uploaded to a server for later analysis and storage.

[1564] This system allows local government employees to carry out their work efficiently and significantly reduces the time it takes to prepare assembly response proposals. In addition, by utilizing past meeting minutes data, it is possible to prepare consistent response proposals.

[1565] The processing flow will be explained below.

[1566] Step 1:

[1567] The user uploads the parliamentary recording (e.g., "session_2023_10.wav") to the server using the interface. The user selects the recording using a web browser and clicks the "Upload" button.

[1568] Step 2:

[1569] The server receives the uploaded recording data. After receiving it, the file is stored in cloud storage, where the file's metadata (e.g., session ID, upload date and time) is also recorded.

[1570] Step 3:

[1571] The server analyzes the received recording data, activates a speech recognition engine, and converts the audio data into text data, which is then stored in a temporary database.

[1572] Step 4:

[1573] The server extracts the text data obtained from the speech recognition engine and uses speaker identification technology to segment it into segments for each speaker, identifying speakers based on, for example, voice patterns or time of day.

[1574] Step 5:

[1575] The server formats the text data of the identified speaker into a format that is easy for the AI ​​model to understand (e.g., JSON format). The formatted data is then stored in cloud storage. The following is an example of the data:

[1576] json

[1577] {

[1578] "session_id": "2023_10",

[1579] "timestamp": "2023-10-15",

[1580] "content": [

[1581] {

[1582] "speaker": "Rep. A",

[1583] "statement": "I'd like to ask about this year's budget."

[1584] },

[1585] {

[1586] "speaker": "Mayor B",

[1587] "statement": "The reason for the budget cuts is..."

[1588] }

[1589] ]

[1590] }

[1591] Step 6:

[1592] The user enters the topic and question for the next session of the Diet on the device. For example, they enter, "I plan to ask a question about the budget at the next session of the Diet." The input is done through a web interface.

[1593] Step 7:

[1594] The server searches for past minutes data based on the input topic, and efficiently extracts related minutes data using metadata.

[1595] Step 8:

[1596] The server uses a generative AI model (e.g., GPT-4) to generate a new draft response based on the extracted minutes data. Below is an example of a generated draft response:

[1597] json

[1598] {

[1599] "suggestions": [

[1600] {

[1601] "topic": "budget",

[1602] "answer": "The budget cuts made last year were made to improve efficiency and were implemented in a way that minimized the impact on the educational environment."

[1603] }

[1604] ]

[1605] }

[1606] Step 9:

[1607] The generated draft answer is displayed on the device, where the user can review it and edit it if necessary. The final draft answer can also be saved in a different format or printed.

[1608] Step 10:

[1609] A user enters a natural language search query, for example, "Show me last year's education budget discussions."

[1610] Step 11:

[1611] The server analyzes the entered search query and searches the database for relevant meeting minutes data, using natural language processing techniques.

[1612] Step 12:

[1613] The search results are displayed on the device. For example, the following search results are displayed on the user's screen:

[1614] text

[1615] Meeting on October 15, 2022:

[1616] Councilman A: "Please explain last year's education budget cuts."

[1617] Mayor B: "The reason behind the budget cuts is..."

[1618] Example 1

[1619] 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."

[1620] Conventional methods for managing parliamentary recording data require a great deal of time and effort for manual minutes creation and data search, resulting in low productivity. It is also difficult to quickly propose consistent content for the next parliamentary session. Therefore, there is a need for a system that can automatically and efficiently create minutes and provide draft answers using generative AI models.

[1621] 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.

[1622] In this invention, the server includes means for receiving recorded data, means for analyzing the received recorded data and converting the audio data into text data, means for formatting the converted text data into a predetermined format, means for saving the formatted text data, means for generating a draft answer using a generative AI model based on previously saved text data, means for outputting the generated draft answer, and means for displaying the output draft answer on a terminal for user input. This makes it possible to automatically and efficiently create minutes and quickly propose consistent draft answers using the generative AI model.

[1623] "Recorded data" refers to data that records in audio format statements and discussions at meetings such as parliaments.

[1624] "Means for receiving" refers to the function or device that allows the server to import the recorded data uploaded by the user.

[1625] "Means for analyzing" refers to the technology or method for analyzing the received recorded data and converting the audio data into text data.

[1626] "Audio data" refers to audio signals stored as recorded data.

[1627] "Text data" is data of character strings that are obtained by analyzing voice data and converting it into linguistic information.

[1628] The "predetermined format" refers to the form or structure used to format the generated text data to ensure consistency and ease of understanding.

[1629] "Storage means" refers to a function or system for storing and managing formatted text data.

[1630] A "generative AI model" refers to a model that uses artificial intelligence to generate responses or suggestions based on specific input data.

[1631] A "draft answer" is a prepared answer that a member of parliament or other relevant person can use to appropriately respond to a question asked in a parliament or other such setting.

[1632] "Means of output" refers to the functions and systems for presenting the generated text data and draft answers to the user.

[1633] A "user input terminal" is a device used by a user to enter data and check results.

[1634] A "natural language search query" refers to a question that a user enters in natural language to search past data.

[1635] "Query-based search means" refers to techniques or methods for retrieving relevant past data from an input natural language search query.

[1636] "Means for outputting search results" refers to the functions and systems for displaying search results to the user.

[1637] "Recording device" refers to a device or equipment for acquiring audio and video data.

[1638] This invention is a system that uses generative AI models to efficiently create parliamentary minutes and automatically generate materials for the next parliamentary session. The system starts with the user uploading audio data, then analyzes the audio data, converts it to text data, stores the data, searches it, and generates draft responses.

[1639] Acquisition of parliamentary recording data

[1640] A user uploads the recorded data to the server via a web interface. For example, after a session ends, the user selects the recorded data (e.g., "session_2023_10.wav") and clicks the upload button. The server receives the data as an HTTP POST request and saves it to cloud storage (e.g., Amazon S3). The server notifies the user when the file has been saved.

[1641] Data analysis and minutes preparation

[1642] The server passes the received recording data to a speech recognition engine, which converts the audio data into text data. This analysis is performed using, for example, the Google Speech-to-Text API. The audio data is converted into a byte stream and sent via the API. The returned text data is split into segments for each speaker using speaker identification technology and formatted into a JSON format that is easy for the generative AI model (e.g., GPT-4) to understand.

[1643] For example, the following JSON data is generated:

[1644] json

[1645] {

[1646] "session_id": "2023_10",

[1647] "timestamp": "2023-10-15",

[1648] "content": [

[1649] {

[1650] "speaker": "Rep. A",

[1651] "statement": "I'd like to ask about this year's budget."

[1652] },

[1653] {

[1654] "speaker": "Mayor B",

[1655] "statement": "The reason for the budget cuts is..."

[1656] }

[1657] ]

[1658] }

[1659] Storage of past meeting minutes data

[1660] The server saves the formatted JSON data in cloud storage. This allows it to be associated with past meeting minutes data. When saving, metadata (e.g., session ID, date) is also saved.

[1661] Generate next parliamentary response

[1662] The user uses an input form on their device to input the topic and questions for the next session. For example, they might input, "I plan to ask a question about the budget at the next session." The server receives this input and searches past minutes data to extract relevant information. It then uses a generative AI model (e.g., GPT-4) to generate new draft responses. An example of this prompt is, "Imagine a question about 'budget cuts' at the next session. Please generate a relevant draft response based on past minutes data." The generated draft response is displayed on the user's device.

[1663] Natural language search function

[1664] Users can search past meeting minutes data by entering natural language search queries into their devices. For example, they can enter "Show me last year's education budget discussions." The server analyzes this input using a natural language processing (NLP) module and searches for relevant data. Search results are displayed on the user's device.

[1665] You'll see specific search results like this:

[1666] text

[1667] Meeting on October 15, 2022:

[1668] Councilman A: "Please explain last year's education budget cuts."

[1669] Mayor B: "The reason behind the budget cuts is..."

[1670] Parliamentary recording and video recording equipment

[1671] Users operate dedicated audio and video recording equipment for council meetings to record speeches and discussions in high quality. The recorded data is later uploaded to a server for analysis and storage. This system enables local government officials to carry out their work efficiently and prepare consistent responses.

[1672] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1673] Step 1: Upload your recordings

[1674] The user uploads the recording (e.g., "session_2023_10.wav") to the server via a web interface. Input: The recording file of the parliament. Output: The uploaded recording file is saved in cloud storage. Specific behavior: The user opens a file selection dialog in a web browser, selects the recording data from the local file system, and presses the upload button. The server receives the HTTP POST request and saves the file in cloud storage.

[1675] Step 2: Analyzing the audio data

[1676] The server retrieves the recorded data from cloud storage and sends it to a speech recognition engine (e.g., Google Speech-to-Text API). Input: The saved recording file. Output: The speech data converted to text data. Specific operation: The server converts the speech data into a byte stream, sends it to the recognition engine, and receives the returned text data.

[1677] Step 3: Formatting the text data

[1678] The server analyzes the acquired text data and uses speaker identification technology to split and format it for each speaker. Input: Text data returned from the speech recognition engine. Output: Meeting minutes data in JSON format. Specific operation: The server analyzes the text data, splits it for each speaker, and generates formatted JSON data.

[1679] Step 4: Save your data

[1680] The server saves the formatted JSON data in cloud storage. Input: Formatted text data (JSON format). Output: Meeting minutes data saved in cloud storage. Specific operation: The server saves the JSON data in the specified directory, along with the metadata.

[1681] Step 5: Generate draft answers for the next session

[1682] The user inputs the topic and question for the next session of the Diet (e.g., "I plan to ask a question about the budget at the next session of the Diet") into the device. Input: The topic and question for the next session of the Diet. Output: Generated draft answers. Specific operation: The server searches past minutes data based on the input topic and question, generates new draft answers using a generative AI model (e.g., GPT-4), and displays them on the device.

[1683] Step 6: Natural Language Search

[1684] A user enters a natural language search query (e.g., "Show me last year's education budget discussions") into a device. Input: Natural language search query. Output: Relevant meeting minutes data. Specific behavior: The server uses a natural language processing (NLP) module to parse the search query, search for relevant data, and display the search results on the device.

[1685] Step 7: Operating the recording equipment

[1686] The user operates audio and video recording equipment specifically designed for the assembly to capture high-quality audio and video data. Input: Data captured by the audio and video recording equipment. Output: The audio and video data is uploaded to the server and stored in cloud storage. Specific operation: The user presses the start recording button on the audio and video recording equipment, which records the entire assembly meeting and then uploads the data to the server.

[1687] (Application example 1)

[1688] 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."

[1689] In logistics centers, creating minutes of meetings and conferences is a time-consuming and labor-intensive process, and labor shortages can lead to inefficiencies. It is also difficult to quickly retrieve information related to the next meeting or task by searching past meeting records. The present invention aims to solve these problems by providing a system that automatically converts meeting recording data into text and generates materials and plans for the next meeting based on a generative AI model.

[1690] 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.

[1691] In this invention, the server includes means for receiving recorded assembly data, means for analyzing the received recorded data and converting the audio data into text data, means for formatting the converted text data into a predetermined format, means for saving the formatted text data, means for generating draft responses using a generative AI model based on previously saved text data, means for inputting the next meeting theme and tasks, means for extracting related information and generating new meeting materials and draft plans, and means for outputting the generated draft responses and meeting materials. This automates the creation of meeting minutes within the logistics center, enabling the efficiency of preparations for the next meeting and tasks.

[1692] "Parliamentary recording data" refers to data files containing recordings of speeches and discussions held in parliaments or meetings.

[1693] "Means for receiving" refers to the function for obtaining recorded data from outside and importing it into the server.

[1694] "Means for analyzing and converting voice data into text data" refers to a function for converting recorded voice data into text information using voice recognition technology.

[1695] "Means for formatting into a specified format" refers to a function for arranging converted text data according to certain rules and formats.

[1696] "Means of storage" refers to the function of storing formatted text data in a database, cloud storage, etc. for future reference or search.

[1697] "Means for generating draft responses using generative AI models" refers to a function that uses AI technology based on past data to automatically create draft responses to be used in the next parliament or meeting.

[1698] "Means for inputting the next meeting topic and tasks" refers to an interface that allows the user to provide the server with information and questions about the next meeting.

[1699] "Means for extracting relevant information and generating new materials and plans" refers to the function of extracting necessary information from past data and using it to create new materials and plans.

[1700] "Means for outputting" refers to a function for providing the generated draft answers and meeting materials to the user.

[1701] "Means for accepting natural language search queries" refers to an interface that allows users to enter search terms in everyday language.

[1702] "Means for searching previously stored text data based on a search query" refers to a function for extracting data that matches the user's input from a database.

[1703] "Means for outputting search results" refers to a function for providing the searched content to the user.

[1704] "Means for additionally analyzing recorded data" refers to a function for further analyzing recorded data within a logistics facility and converting it into text data, etc.

[1705] "Function to record meeting recording data" refers to the function to collect the contents of meetings and assemblies in audio format.

[1706] "The function of recording recorded data" refers to the function of collecting the contents of meetings and assemblies in video format.

[1707] "Recording and video recording equipment" refers to devices and equipment used to record the contents of meetings and assemblies as audio or video.

[1708] MODE FOR CARRYING OUT THE INVENTION

[1709] This system automatically analyzes recorded meeting data at a logistics center, converts it into text data, and saves it. It can also automatically generate materials and plans for the next meeting using a generative AI model. It also has the ability to search past meeting records using natural language search queries and provide the results to users.

[1710] System Overview

[1711] The system consists of the following main components:

[1712] 1. Means of receiving parliamentary recordings

[1713] 2. Means of converting audio data into text data

[1714] 3. A means of formatting text data into a specified format

[1715] 4. Means of saving text data

[1716] 5. How to generate draft responses using generative AI models

[1717] 6. A way to enter the next meeting topic or task

[1718] 7. A means to extract relevant information and generate new meeting materials and plans

[1719] 8. Means for outputting the generated response plan and meeting materials

[1720] 9. A way to accept natural language search queries

[1721] 10. A way to search past data based on a search query

[1722] 11. How to output search results

[1723] 12. Additional means of analyzing recording data

[1724] 13. Function to record conference recording data

[1725] 14. Audio and video recording devices with the function of recording video data

[1726] System Operation

[1727] The server receives the parliamentary recording data, analyzes it, and converts the audio data into text data. This speech recognition uses technologies such as DeepSpeech and Google Speech-to-Text API. The converted text data is formatted and stored in cloud storage (e.g., AWS S3).

[1728] Users can input the topic or task for the next meeting into their device. For example, they can input a prompt such as, "Please create a document to discuss the logistics situation at the next meeting." The server then searches past text data and extracts relevant information. Based on this extracted information, a generative AI model (e.g., GPT-4) is used to generate new meeting materials and draft plans. The generated materials and draft responses are then displayed on the user's device.

[1729] Additionally, users can search for past meeting records using natural language search queries, such as a prompt such as "Show me meeting records from last year regarding logistics status." The server analyzes the query, searches for relevant data, and provides the results to the user.

[1730] Hardware and software used

[1731] The system uses the following hardware and software:

[1732] Server: Amazon Web Services (AWS) EC2

[1733] Cloud storage: AWS S3

[1734] Speech recognition engine: DeepSpeech, Google Speech-to-Text API

[1735] Generative AI model: OpenAI GPT-4

[1736] Natural language processing libraries: SpaCy, NLTK

[1737] Specific examples

[1738] For example, suppose a logistics center manager holds a weekly meeting and records it. The recording is then uploaded to a server, which analyzes the data and creates meeting minutes. For next week's meeting, if a user types into their device, "Please generate this week's plan based on last week's minutes," the server will search past data and automatically generate a new plan using a generative AI model. This can improve work efficiency.

[1739] Examples of prompts include "Please prepare materials to talk about the logistics situation at the next meeting" and "Show me the minutes of last year's meeting about the logistics situation."

[1740] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1741] Step 1:

[1742] A user acquires parliamentary recording data using audio and video recording equipment and uploads the data to the server. In this example, the recording file (e.g., "meeting_2023_10.wav") is sent to the server via a web interface. The input is an audio file, and the output is an audio file stored in cloud storage.

[1743] Step 2:

[1744] The server analyzes the received recording data and converts the audio data into text data. This process uses a speech recognition engine (DeepSpeech or Google Speech-to-Text API). The input is an audio file, and the output is the converted text data.

[1745] Step 3:

[1746] The server formats the converted text data into a specified format. Specifically, it uses speaker identification technology to split the text data for each speaker and converts it into JSON format. The input is raw text data, and the output is formatted text data.

[1747] Step 4:

[1748] The server saves the formatted text data in cloud storage (AWS S3), along with metadata (e.g., session ID, date). The input is the formatted text data, and the output is the data saved in cloud storage.

[1749] Step 5:

[1750] The user inputs the next meeting topic and task into the terminal. For example, the user might input, "Please prepare materials to talk about the logistics situation at the next meeting." The input is a prompt text, and the output is sending the prompt text to the server.

[1751] Step 6:

[1752] The server searches for information related to the prompt text entered by the user based on previously saved text data. The input is the prompt text and past text data, and the output is the searched related information.

[1753] Step 7:

[1754] Based on the related information retrieved by the server, a generative AI model (GPT-4) is used to generate meeting materials and plan proposals. The input is related information, and the output is new meeting materials and plan proposals.

[1755] Step 8:

[1756] The server outputs the generated meeting materials and plans to the terminals to provide them to the users. The input is the new meeting materials and plans, and the output is the materials and plans displayed on the user's terminal.

[1757] Step 9:

[1758] A user inputs a search query in natural language into a terminal. For example, "Show me the minutes of the meeting about the logistics situation last year." The input is a search query in natural language, and the output is a search query sent to the server.

[1759] Step 10:

[1760] The server searches past text data based on the search query. Here, it references a database of past meeting minutes and extracts relevant data. The input is the search query and past text data, and the output is the search result data.

[1761] Step 11:

[1762] The server outputs the search results to the terminal, allowing the user to view the results. The input is the search result data, and the output is the search result displayed on the user's terminal.

[1763] 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.

[1764] The system of the present invention uses a generative AI model to create minutes of parliamentary sessions and automatically generates the next parliamentary session materials (such as draft parliamentary responses) based on that data. Furthermore, by combining it with an emotion engine that recognizes the user's emotions, it is possible to generate more appropriate draft responses that reflect the user's intentions and emotions. Below, we will explain in detail how this system is implemented.

[1765] Acquisition of parliamentary recording data

[1766] The user uploads the parliamentary recording (e.g., "session_2023_10.wav") to the server using the interface. The user selects the recording via a web browser and clicks the "Upload" button. The file is received by the server and stored in cloud storage.

[1767] Data analysis and minutes preparation

[1768] The server analyzes the uploaded recording data, activates a speech recognition engine, and converts the audio data into text data. The speech recognition technology used is DeepSpeech or the Google Speech-to-Text API, for example. The converted text data is then stored in a temporary database.

[1769] Speaker Identification and Data Shaping

[1770] The server extracts the text data obtained from the speech recognition engine and divides it into segments for each speaker using speaker identification technology. For example, speakers can be identified based on voice patterns or time of day. The server then formats the text data of the identified speaker into a format (e.g., JSON) that is easy for the generative AI model to understand. The formatted data is then saved in cloud storage. The following is an example of the data:

[1771] json

[1772] {

[1773] "session_id": "2023_10",

[1774] "timestamp": "2023-10-15",

[1775] "content": [

[1776] {

[1777] "speaker": "Rep. A",

[1778] "statement": "I'd like to ask about this year's budget."

[1779] },

[1780] {

[1781] "speaker": "Mayor B",

[1782] "statement": "The reason for the budget cuts is..."

[1783] }

[1784] ]

[1785] }

[1786] Generating Answer Plans and Using an Emotion Engine

[1787] The user inputs the topic and questions for the next session of the Diet on the terminal. For example, the user might input, "I plan to ask a question about the budget at the next session of the Diet." The input is done via a web interface. The server searches past minutes data based on this input and extracts relevant information.

[1788] The server uses a generative AI model (e.g., GPT-4) to generate new response proposals based on the extracted information. Furthermore, an emotion engine recognizes the user's emotional data and adjusts the response proposals based on that data. For example, if the user is recognized as "nervous," the response proposals will include expressions that convey a more reassuring feeling.

[1789] json

[1790] {

[1791] "suggestions": [

[1792] {

[1793] "topic": "budget",

[1794] "answer": "The budget cuts made last year were made to improve efficiency and we have taken measures to minimize the impact on the educational environment. Please rest assured."

[1795] }

[1796] ]

[1797] }

[1798] Storing and using emotional data

[1799] The server stores the user's emotional data recognized by the emotion engine in cloud storage. The stored emotional data is used to generate future parliamentary response drafts. This results in more consistent response drafts and increases user confidence.

[1800] Natural language search function

[1801] A user can search the database of past meeting minutes using a natural language search query. For example, they can type "Show me last year's education budget discussion." The server analyzes this natural language query and searches for relevant meeting minutes data. As a result, the following search results are displayed on the user's device:

[1802] text

[1803] Meeting on October 15, 2022:

[1804] Councilman A: "Please explain last year's education budget cuts."

[1805] Mayor B: "The reason behind the budget cuts is..."

[1806] Parliamentary recording and video recording equipment

[1807] Users can record high-quality audio and video of parliamentary speeches and debates using dedicated parliamentary audio and video recording equipment, which will then be uploaded to a server for later analysis and storage.

[1808] This system allows local government employees to carry out their work efficiently and significantly reduces the time it takes to prepare assembly response drafts. Furthermore, by taking into consideration the user's feelings, it is possible to provide more reliable response drafts. In addition, by utilizing past minutes data, it is possible to prepare consistent response drafts.

[1809] The processing flow will be explained below.

[1810] Step 1:

[1811] The user uploads the parliamentary recording (e.g., "session_2023_10.wav") to the server using the interface. The user selects the recording via a web browser and clicks the "Upload" button.

[1812] Step 2:

[1813] The server receives the uploaded recording data. After receiving it, the file is stored in cloud storage, where the file's metadata (e.g., session ID, upload date and time) is also recorded.

[1814] Step 3:

[1815] The server analyzes the received recording, invokes a speech recognition engine (e.g., DeepSpeech or Google Speech-to-Text API), and converts the audio data into text data, which is then stored in a temporary database.

[1816] Step 4:

[1817] The server extracts the text data obtained from the speech recognition engine and uses speaker identification technology to segment it into segments for each speaker, identifying speakers based on, for example, voice patterns or time of day.

[1818] Step 5:

[1819] The server formats the text data of the identified speaker into a format that is easy for the AI ​​model to understand (e.g., JSON format). The formatted data is then stored in cloud storage. The following is an example of the data:

[1820] json

[1821] {

[1822] "session_id": "2023_10",

[1823] "timestamp": "2023-10-15",

[1824] "content": [

[1825] {

[1826] "speaker": "Rep. A",

[1827] "statement": "I'd like to ask about this year's budget."

[1828] },

[1829] {

[1830] "speaker": "Mayor B",

[1831] "statement": "The reason for the budget cuts is..."

[1832] }

[1833] ]

[1834] }

[1835] Step 6:

[1836] The user inputs the topic and questions for the next council meeting on the terminal. For example, the user inputs, "I plan to ask a question about the budget at the next council meeting." The input is done via a web interface.

[1837] Step 7:

[1838] The server searches for past minutes data based on the input topic, and efficiently extracts related minutes data using metadata.

[1839] Step 8:

[1840] The server uses a generative AI model (e.g., GPT-4) to generate a new draft response based on the extracted minutes data. Below is an example of a generated draft response:

[1841] json

[1842] {

[1843] "suggestions": [

[1844] {

[1845] "topic": "budget",

[1846] "answer": "The budget cuts made last year were made to improve efficiency and were implemented in a way that minimized the impact on the educational environment."

[1847] }

[1848] ]

[1849] }

[1850] Step 9:

[1851] The server uses the emotion engine to recognize the user's emotion, for example, the emotion engine recognizes that the user is "nervous."

[1852] Step 10:

[1853] The server reflects the emotional data it recognizes in the generated answer proposals. For example, if the server recognizes that the user is "nervous," the answer proposals include expressions that make the user feel more at ease.

[1854] Step 11:

[1855] The generated draft answer is displayed on the device, where the user can review it and edit it if necessary. The final draft answer can also be saved in a different format or printed.

[1856] Step 12:

[1857] A user enters a natural language search query, for example, "Show me last year's education budget discussions."

[1858] Step 13:

[1859] The server analyzes the entered search query and searches the database for relevant meeting minutes data, using natural language processing techniques.

[1860] Step 14:

[1861] The search results are displayed on the device. For example, the following search results are displayed on the user's screen:

[1862] text

[1863] Meeting on October 15, 2022:

[1864] Councilman A: "Please explain last year's education budget cuts."

[1865] Mayor B: "The reason behind the budget cuts is..."

[1866] Example 2

[1867] 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."

[1868] The task of creating minutes of assembly meetings and response materials for the next assembly meeting requires time and effort, placing a heavy burden on staff. Furthermore, creating response materials requires appropriately reflecting the content of past discussions while also taking into account the user's emotions and intentions, but there is no efficient way to do this. In conventional technologies, creating minutes and generating response materials is often done manually, resulting in reduced work efficiency and inconsistent quality. This invention is intended to solve these problems.

[1869] The specific processing by the specific processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for receiving parliamentary recording data, means for analyzing the received recording data and converting the audio data into text data, means for formatting the converted text data into a predetermined data format, means for generating a draft response using a generative AI model based on previously saved text data, means for reflecting the user's emotions in the generated draft response using an emotion engine, and means for outputting the generated draft response. This automates the creation of minutes and the generation of response materials, improving work efficiency and ensuring consistent quality. Furthermore, because response materials are generated taking the user's emotions into consideration, more appropriate and reliable materials can be created.

[1870] "Parliamentary recording data" refers to audio data of speeches and discussions in parliament.

[1871] A "generative AI model" refers to an artificial intelligence model that generates new outputs based on input data.

[1872] "Emotion engine" refers to a program or function that recognizes a user's emotions and analyzes the emotion data.

[1873] "Draft answers" refers to prepared answers to questions asked in parliament.

[1874] "Converted text data" refers to character data converted from voice data using voice recognition technology.

[1875] "Data format" refers to the format or standard for organizing and arranging data in a certain structure.

[1876] A "natural language search query" refers to a search request made to a database using natural language.

[1877] "Sound recording equipment" means devices for recording sound and video.

[1878] "Cloud storage" refers to an online storage service that stores and manages data via the Internet.

[1879] A "server" refers to a computer system that provides services and data to other computers over a network.

[1880] "Terminal" refers to a device used by a user for input and display.

[1881] "Text data" refers to data expressed as character information.

[1882] The system of the present invention utilizes a generative AI model and an emotion engine to create minutes of parliamentary meetings and automatically generate materials for responses to questions at the next parliamentary meeting. A detailed embodiment of this system will be described.

[1883] Overall system overview

[1884] This system analyzes recordings of parliamentary sessions, converts the audio data into text data, and uses that text data to automatically generate draft responses for the next session. It can also recognize the user's emotions and generate draft responses that reflect those emotions. This system consists of three main components: a server, a terminal, and a user.

[1885] Hardware and Software Configuration

[1886] The server receives the recorded data and analyzes the audio data. It is equipped with a high-performance processor and a large amount of memory. The speech recognition engine uses common speech recognition software such as DeepSpeech or Google Speech-to-Text API. The server also converts the analyzed text data into a specified data format and stores it in a temporary database.

[1887] The terminal provides an interface for users to input the topic and questions for the next session. This terminal is a standard PC or tablet and is connected to the server via a web browser. Users can use the terminal to upload recording data as well as check the generated response plans.

[1888] Users upload the recorded data of assembly meetings to the server using their devices. They input the topic and questions of the next assembly meeting, and check and use the proposed answers provided by the server. Users can also input search queries to search for past minutes data.

[1889] Processing flow and specific examples

[1890] The system begins by a user uploading a recording of a parliamentary session. The recording (e.g., "session_2023_10.wav") is then stored in cloud storage by the server. The server then activates a speech recognition engine, analyzes the recording, and converts the audio data into text data. This text data is then split by speaker using speaker identification technology and formatted into a format that is easy for the generative AI model to understand.

[1891] For example, a statement such as "I'd like to ask about this year's budget" is linked to "Council Member A," and a statement such as "The reason for the budget cuts is..." is linked to "Mayor B." This results in the data being formatted as text like this:

[1892] Session ID: 2023_10

[1893] Timestamp: 2023-10-15

[1894] Contents:

[1895] Councilman A: "I'd like to ask about this year's budget."

[1896] Mayor B: 'The reason for the budget cuts is...'

[1897] "

[1898] Next, the user inputs the next session's theme and question on the device. For example, they might input, "I plan to ask a question about the budget at the next session." The server extracts relevant information from past minutes data and uses this information to input prompts into the generative AI model to generate new draft responses. The emotion engine also recognizes the user's emotions, and if it recognizes that the user is "nervous," the generated draft responses will include expressions that convey a sense of security.

[1899] Examples of prompts are:

[1900] "Parliamentary Data:

[1901] Session ID: 2023_10

[1902] Timestamp: 2023-10-15

[1903] Contents:

[1904] Councilman A: "I'd like to ask about this year's budget."

[1905] Mayor B: 'The reason for the budget cuts is...'

[1906] Next Congress Theme and User Sentiments:

[1907] Theme: 'Budget'

[1908] Question: I plan to ask a question about the budget at the next Congress. I would like to know specifically why the budget cuts were made.

[1909] Emotion: 'Nervous'

[1910] Generated answer plan:

[1911] Topic: 'Budget'

[1912] Response: 'The budget cuts made last year were made to improve efficiency and we have taken measures to minimize the impact on the educational environment. Please rest assured.'

[1913] "

[1914] The server saves the generated answer plan in cloud storage and displays it on the user's device. The user's emotional data is also saved in cloud storage and used to generate answer plans from the next time onwards, making it possible to create consistent answer plans.

[1915] The above is a specific embodiment of the system. This system automates the creation of assembly minutes and the generation of response materials for the next assembly meeting, and can provide high-quality materials that take user emotions into consideration.

[1916] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1917] Step 1: Obtaining Congressional Recording Data

[1918] Input: The user uploads a recording file (e.g., "session_2023_10.wav") through the designated interface.

[1919] How it works: A user uses a web browser to select a recording and presses the "upload" button.

[1920] Output: The server saves this recording to cloud storage and records the file's metadata (e.g., file name, upload date and time).

[1921] Step 2: Analyze the data and create a transcript

[1922] Input: Recording data files stored in cloud storage.

[1923] How it works: The server launches a speech recognition engine (e.g., DeepSpeech or Google Speech-to-Text API) and reads the recording. The server converts the audio data into text. This conversion process converts what is being said into text format.

[1924] Output: The converted text data is saved in a temporary database.

[1925] Step 3: Speaker identification and data conditioning

[1926] Input: Text data stored in a temporary database.

[1927] How it works: The server uses speaker identification technology to identify each speaker, using characteristics such as voice patterns and time of day. It then organizes the text data for each identified speaker and formats it into a data format (e.g., JSON) that is easy for the generative AI model to understand.

[1928] Output: The formatted text data is saved to cloud storage.

[1929] Step 4: Enter the theme and questions for the next assembly meeting

[1930] Input: The user enters the topic or question for the next session through a web interface (e.g., "I have a question about the budget for the next session").

[1931] How it works: The user uses a terminal to enter the required information and send it to the server.

[1932] Output: The server receives this input data and prepares it for the next processing step.

[1933] Step 5: Generate a response and use the emotion engine

[1934] Input: User questions and past meeting minutes data.

[1935] How it works: The server extracts relevant information from past meeting minutes data and inputs prompt sentences into a generative AI model. A generative AI model (e.g., GPT-4) is used to generate new response proposals. The emotion engine then analyzes the user's emotion data and adjusts the response proposals based on the recognized emotion. For example, "If the user is nervous, add expressions that will make them feel at ease."

[1936] Output: A response plan reflecting the sentiment is generated and saved in cloud storage.

[1937] Step 6: Outputting the generated answer plan

[1938] Input: Response plan data stored on the server.

[1939] Operation: The server sends the generated answer plan to the user's terminal, where the user can view the answer plan.

[1940] Output: The proposed answer displayed on the user's device.

[1941] Step 7: Storing and using emotion data

[1942] Input: User emotion data analyzed by the emotion engine.

[1943] How it works: The server stores the emotional data in cloud storage. This data is used to understand the user's emotional tendencies for future response generation.

[1944] Output: Emotion data stored in cloud storage.

[1945] Step 8: Processing Natural Language Search Queries

[1946] Input: A natural language search query entered by a user through a web interface (e.g., "Show me last year's education budget discussions").

[1947] How it works: The server analyzes the search query and searches for relevant information from past meeting minutes data.

[1948] Output: Search results displayed on the user's device.

[1949] In this way, each processing step works together to automate the entire process, from creating minutes of parliamentary sessions to generating response materials for the next session of parliament.

[1950] (Application example 2)

[1951] 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."

[1952] Conventional meeting minutes creation systems simply convert the minutes into text and store them, but there is a problem in that it takes a lot of time and effort to properly prepare for the next meeting and generate draft responses. It is also difficult to create draft responses that fully reflect the user's emotions and intentions, which can lead to a decrease in user satisfaction. Furthermore, there is a demand for similar technology to be applied to other industries, and in the food delivery field in particular, there is a strong need for technology that combines review analysis and emotion recognition to improve customer satisfaction.

[1953] 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.

[1954] In this invention, the server includes means for receiving parliamentary recording data, means for analyzing the received recording data and converting the audio data into text data, means for formatting the converted text data into a predetermined format, means for saving the formatted text data, means for generating draft responses using a generative AI model based on previously saved text data, means for analyzing a user's emotional data using an emotion recognition engine and adjusting the draft responses using the generative AI model based on this, means for generating personalized messages and suggestions that reflect the user's emotional data at the time of the next order, and means for outputting the generated draft responses. This makes it possible to automatically generate appropriate draft responses and suggestions that are linked to the user's emotions.

[1955] "Parliamentary recording data" refers to audio recordings of speeches and discussions in parliament.

[1956] "Means" refers to a method or apparatus for performing a specific function or process in a patent.

[1957] "Audio data" refers to digitalized sound information acquired through a voice input device such as a microphone.

[1958] "Text data" is a form of data expressed as character information.

[1959] A "format" is a set of rules for arranging data into a particular structure or form.

[1960] A "generative AI model" is an algorithm or program that uses machine learning technology to create new information.

[1961] A "draft reply" is a proposed response to a specific question or topic in parliament or elsewhere.

[1962] An "emotion recognition engine" is a technology that analyzes and identifies human emotions from data such as voice and text.

[1963] "User" refers to a person or organization that uses the system.

[1964] A "message" is a sentence or text intended to convey specific information.

[1965] A "suggestion" is an idea or opinion submitted recommending a particular action or option.

[1966] A "system" is a collection of integrated structures or devices in which multiple elements work together to perform a specific task.

[1967] System Overview

[1968] This system applies technology for analyzing recorded conference data to generate new suggestions and messages based on users' emotional data in order to improve customer satisfaction in food delivery services. The system allows users to upload audio reviews, analyzes the audio data to recognize emotions, and provides personalized messages and suggestions for the next order. The entire system consists of a server, a terminal, and a user.

[1969] Hardware and software used

[1970] Speech Recognition: DeepSpeech, Google Speech-to-Text API

[1971] Generative AI model: GPT-4

[1972] Emotion recognition engine: IBM Watson Tone Analyzer

[1973] Cloud storage: Amazon S3

[1974] Database: MySQL, Firebase

[1975] Web interface: React.js, Node.js

[1976] Processing Details

[1977] 1. Review upload function

[1978] After ordering from a food delivery service, users record their impressions in audio format and upload them to the app. For example, a user might upload a comment such as, "The food I ordered was delicious, but the delivery was slow."

[1979] 2. Review Analysis and Emotion Recognition

[1980] The server receives the audio files uploaded by users and stores them in Amazon S3. It then converts the audio data into text using DeepSpeech or the Google Speech-to-Text API and stores it in a MySQL database. The converted text data undergoes sentiment analysis using IBM Watson Tone Analyzer to identify the user's emotional data.

[1981] 3. Proposal generation for next order

[1982] The next time the user logs in to the app and attempts to place an order, the server generates a prompt based on the stored emotional and text data. For example, "Based on the user's review, the last order was delayed but the user was satisfied with the taste. Generate a message with highlighted menu items and a discount for the next order based on this information." This prompt is input into GPT-4, which generates an appropriate message and suggestions. The generated message and suggestions are then displayed on the user's order screen.

[1983] Specific examples

[1984] For example, if a user leaves a review about their last order saying, "The food was delicious, but the delivery was late and I was a little disappointed," the next time they order, they might see a suggestion like this: "We apologize for the delay with your last order. To ensure we can deliver your order on time, please take a look at our new Chef's Specials. Plus, we'll give you a 10% discount on your next order."

[1985] How the program is implemented

[1986] The server uses the above-mentioned methods to analyze voice data, generate emotion recognition data, generate prompts based on the stored data, and output messages and suggestions using a generative AI model. This system makes it possible to automatically generate appropriate responses and suggestions that are linked to the user's emotions.

[1987] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1988] Step 1:

[1989] A user uses a food delivery service and records an audio review after placing an order. The user opens the application, uses the recording function to record their review, and uploads the audio file (e.g., "foodie_review_2023_10_01.wav") to the app. The input is the user's voice data, which is then stored in cloud storage.

[1990] Step 2:

[1991] The server receives the audio file uploaded by the user and stores it in cloud storage (Amazon S3). Next, the server invokes a speech recognition engine (DeepSpeech or Google Speech-to-Text API) to convert the audio data into text data. In this case, the input is the audio file and the output is text data. The converted text data (e.g., "The food was delicious, but the delivery was slow") is stored in a temporary database (MySQL).

[1992] Step 3:

[1993] The server uses an emotion recognition engine (IBM Watson Tone Analyzer) to analyze the converted text data and recognize the user's emotional data. In this process, the input is text data, and the output is the user's emotional data (e.g., "dissatisfied" or "satisfied"). The analysis results are stored in cloud storage or a database.

[1994] Step 4:

[1995] The next time the user logs into the app and attempts to place a new order, the server retrieves past reviews and sentiment data. The server generates a prompt based on this data and sends it to a generative AI model (GPT-4) to generate an appropriate message or suggestion. The input is past text data and sentiment data, and the output is the generated message or suggestion. An example of a prompt is, "According to the user's review, the food was delayed on their last order, but they were satisfied with the taste. Based on this information, please generate a message including featured menu items and discounts for their next order."

[1996] Step 5:

[1997] The generative AI model (GPT-4) generates new suggestions and messages based on the prompt. For example, "We apologize for the delay in your last order. As a special favor, we invite you to try our new menu item. We'll give you a 10% discount on your next order." The input is the prompt, and the output is a personalized message.

[1998] Step 6:

[1999] The server displays the generated message on the user's order screen. The user can then place a new order based on the displayed message and suggestions. The input is the generated message, and the output is the user's display screen. This improves the user experience and increases ordering satisfaction.

[2000] 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.

[2001] 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.

[2002] 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.

[2003] 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.

[2004] FIG. 9 is a diagram illustrating 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 actions 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.

[2005] 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.

[2006] 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).

[2007] 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.

[2008] 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."

[2009] 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.

[2010] 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).

[2011] 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.

[2012] 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.

[2013] 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.

[2014] 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.

[2015] 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.

[2016] 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.

[2017] 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.

[2018] 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.

[2019] 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.

[2020] 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.

[2021] The following is further disclosed regarding the above embodiment.

[2022] (Claim 1)

[2023] a means for receiving the parliamentary recording;

[2024] means for analyzing the received recorded data and converting the audio data into text data;

[2025] means for formatting the converted text data into a predetermined format;

[2026] a means for storing the formatted text data;

[2027] A means for generating draft responses using a generative AI model based on previously stored text data;

[2028] A means for outputting the generated answer plan;

[2029] A system including:

[2030] (Claim 2)

[2031] a means for accepting natural language search queries;

[2032] a means for searching previously stored text data based on a search query;

[2033] a means for outputting search results;

[2034] The system of claim 1 further comprising:

[2035] (Claim 3)

[2036] 2. The system of claim 1, further comprising a parliamentary recording and video recording device having a function of recording parliamentary recording data and a function of recording video data.

[2037] "Example 1"

[2038] (Claim 1)

[2039] means for receiving the recording data;

[2040] means for analyzing the received recorded data and converting the audio data into text data;

[2041] means for formatting the converted text data into a predetermined format;

[2042] a means for storing the formatted text data;

[2043] A means for generating draft responses using a generative AI model based on previously stored text data;

[2044] A means for outputting the generated answer plan;

[2045] a means for displaying the outputted answer plan on a terminal where the user inputs the answer;

[2046] A system including:

[2047] (Claim 2)

[2048] a means for accepting natural language search queries;

[2049] a means for searching previously stored text data based on a search query;

[2050] a means for outputting search results;

[2051] means for displaying search results on the terminal on which the user inputs the results;

[2052] The system of claim 1 further comprising:

[2053] (Claim 3)

[2054] 10. The system of claim 1, further comprising a recording device having a function of recording sound data and a function of recording video data.

[2055] "Application Example 1"

[2056] (Claim 1)

[2057] a means for receiving the parliamentary recording;

[2058] means for analyzing the received recorded data and converting the audio data into text data;

[2059] means for formatting the converted text data into a predetermined format;

[2060] a means for storing the formatted text data;

[2061] A means for generating draft responses using a generative AI model based on previously stored text data;

[2062] A way to input the next meeting topic and tasks,

[2063] A means for extracting relevant information to generate new discussion materials and proposals;

[2064] a means for outputting the generated draft answers and meeting materials;

[2065] A system including:

[2066] (Claim 2)

[2067] a means for accepting natural language search queries;

[2068] a means for searching previously stored text data based on a search query;

[2069] a means for outputting search results;

[2070] A means for additionally analyzing recorded data within the logistics facility;

[2071] The system of claim 1 further comprising:

[2072] (Claim 3)

[2073] 10. The system according to claim 1, further comprising a recording / video recording device having a function of recording the conference recording data and a function of recording the video recording data.

[2074] "Example 2: Combining Emotion Engines"

[2075] (Claim 1)

[2076] a means for receiving the parliamentary recording;

[2077] means for analyzing the received recorded data and converting the audio data into text data;

[2078] means for formatting the converted text data into a predetermined data format;

[2079] a means for storing the formatted text data;

[2080] A means for generating draft responses using a generative AI model based on previously stored text data;

[2081] A means for reflecting the user's emotions on the generated answer plan using an emotion engine;

[2082] A means for outputting the generated answer plan;

[2083] A system including:

[2084] (Claim 2)

[2085] a means for accepting natural language search queries;

[2086] a means for searching previously stored text data based on a search query;

[2087] a means for outputting search results;

[2088] The system of claim 1 further comprising:

[2089] (Claim 3)

[2090] 10. The system of claim 1, further comprising a recording device having a function of recording the parliamentary recording data and a function of recording the video data.

[2091] "Application example 2 when combining emotion engines"

[2092] (Claim 1)

[2093] a means for receiving the parliamentary recording;

[2094] means for analyzing the received recorded data and converting the audio data into text data;

[2095] means for formatting the converted text data into a predetermined format;

[2096] a means for storing the formatted text data;

[2097] A means for generating draft responses using a generative AI model based on previously stored text data;

[2098] A means of analyzing the user's emotional data using an emotion recognition engine and adjusting the answer plan using a generative AI model based on this data;

[2099] A means for generating personalized messages and suggestions that reflect the user's emotional data when placing a next order;

[2100] A means for outputting the generated answer plan;

[2101] A system including:

[2102] (Claim 2)

[2103] a means for accepting natural language search queries;

[2104] a means for searching previously stored text data based on a search query;

[2105] a means for outputting search results;

[2106] The system of claim 1 further comprising:

[2107] (Claim 3)

[2108] 2. The system of claim 1, further comprising a parliamentary recording and video recording device having a function of recording parliamentary recording data and a function of recording video data. [Explanation of symbols]

[2109] 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. a means for receiving parliamentary recording data; means for analyzing the received recorded data and converting the audio data into text data; means for formatting the converted text data into a predetermined format; a means for storing the formatted text data; A means for generating draft responses using a generative AI model based on previously stored text data; A means for outputting the generated answer plan; A system including:

2. a means for accepting natural language search queries; a means for searching previously stored text data based on a search query; a means for outputting search results; The system of claim 1 further comprising:

3. The system according to claim 1, further comprising a conference recording and video recording device having a function of recording conference recording data and a function of recording video data.

Images