System

Abstract

A system is provided.SOLUTION: A system comprising: means for converting call content into text in real time; means for analyzing the call content converted into text using a generation model and detecting a suspicious conversation; and means for automatically reporting to a fraud report window when a suspicious conversation is detected.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] As the aging of society progresses, the number of fraud cases is increasing and the methods are becoming more sophisticated. Fraud over the phone is particularly serious, and many methods are seen that target elderly people and trick them into making deposits. Conventional countermeasures against this have limited effectiveness, so there is a need for a method to detect fraud in real time and prevent it before it happens. In particular, a system is needed that can deal with complex fraud methods that are difficult for humans to judge. [Means for solving the problem]

[0005] This invention provides a system that converts phone conversations into text in real time and analyzes the converted text using a generative model. When a suspicious conversation is detected, it automatically reports it to a fraud hotline, preventing fraud before it occurs. It also has the effect of deterring fraudulent activity by playing an announcement at the start of a call that states, "This conversation is being recorded." Furthermore, by providing a means to manage a list that excludes conversations with specific callers from detection, it is possible to implement effective fraud prevention measures without interrupting important calls.

[0006] "Call content" refers to the content of speech exchanged over the telephone, that is, voice data.

[0007] "Real-time" refers to processing at the present time, meaning processing immediately without delay.

[0008] "Text conversion" is the process of converting audio data into a string of characters.

[0009] "Generative model" refers to an algorithm that uses generative adversarial networks (GANs) and other machine learning to analyze and classify data.

[0010] "Analysis" refers to the process of analyzing the text of a call and understanding its context.

[0011] "Suspicious conversations" are those that contain statements or phrases that are deemed to be potentially fraudulent.

[0012] "Reporting Line" refers to an agency or service that receives specific information to prevent fraud.

[0013] "Automatic" refers to an action performed by a machine or program without human intervention.

[0014] "Fraud" refers to the act of taking another person's property by dishonest or malicious means.

[0015] "Deterrence" refers to preventative measures taken to prevent fraudulent activity from occurring.

[0016] An "exclusion list" refers to a list used to exclude items that meet certain conditions from being detected. [Brief explanation of the drawings]

[0017] [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

[0018] 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.

[0019] First, the terms used in the following description will be explained.

[0020] 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).

[0021] 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.

[0022] 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.

[0023] 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.

[0024] 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."

[0025] [First embodiment]

[0026] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0027] 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.

[0028] 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).

[0029] 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.

[0030] 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.

[0031] 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.

[0032] 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.

[0033] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0034] 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.

[0035] 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.

[0036] 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.

[0037] 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."

[0038] The present invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model. Specifically, it detects suspicious conversations and automatically reports them to a fraud hotline as necessary, thereby preventing fraud damage. The system's program processing is described in detail below.

[0039] System Overview

[0040] In this system, when a user initiates a call, the device collects voice data, and the server converts the voice data into text in real time. In addition, the server uses a generative model to analyze the text of the call and detect suspicious conversations. If a suspicious conversation is detected, the server automatically reports it to a fraud hotline, and the necessary measures are taken.

[0041] Announcement at the start of a call

[0042] When a user starts a call, the device plays an announcement saying, "This conversation is being recorded." This announcement acts as a deterrent to fraudulent activity on the other end of the call.

[0043] Speech to text

[0044] During a call, voice data is collected by the user's device and sent to a server, which then uses a voice recognition service to convert the voice data into text in real time, thereby recording the contents of the call as text information.

[0045] Text Analysis

[0046] The server inputs the real-time text of the call into a generative model for analysis. The generative model has an algorithm that detects specific keywords and phrases and identifies suspicious conversations. For example, if keywords such as "deposit," "PIN number," or "transfer" are detected, it will determine that there is a possibility of fraud.

[0047] Automatic reporting of suspicious conversations

[0048] If the generative model detects a suspicious conversation, the server automatically reports it to a fraud hotline. The report includes the content of the call and the generated warning message. The hotline then contacts the user, confirms the situation, and takes any necessary measures.

[0049] Managing the Exclusion List

[0050] Users can also exclude specific callers from detection. The device has an exclusion list management function that allows users to add the IDs of specific callers (for example, family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system, so the system can be operated without interfering with important calls.

[0051] Specific examples

[0052] Consider a case where an elderly user receives a fraudulent phone call, claiming to be an "urgent call from the bank," and is tricked into making a deposit. When the call begins, the device announces, "This conversation is being recorded," and audio data is collected. This audio data is converted into text in real time and analyzed by a generative model on a server. If the generative model detects keywords such as "deposit" and "bank" and determines that the conversation is suspicious, the server automatically reports the incident to a fraud hotline. The hotline contacts the user, confirms the situation, and takes necessary measures to prevent fraud before it occurs.

[0053] In this way, the present invention provides a system that can detect suspicious conversations in real time and respond quickly, thereby effectively preventing fraud against elderly people and others.

[0054] The processing flow will be explained below.

[0055] Step 1:

[0056] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," to deter the other party from engaging in fraudulent activity.

[0057] Step 2:

[0058] The device collects the audio data and sends the audio stream to the server. The audio stream is collected in real time to cover the entire call.

[0059] Step 3:

[0060] The server receives the audio stream and uses a speech recognition service to convert the audio data into text in real time, resulting in a written record of the conversation.

[0061] Step 4:

[0062] The server inputs the transcribed conversations into a generative model, which uses natural language processing (NLP) algorithms to analyze the text and detect specific keywords and phrases.

[0063] Step 5:

[0064] The generative model detects suspicious conversations. For example, if a conversation contains keywords such as "deposit," "PIN number," or "transfer," it determines that it may be fraudulent.

[0065] Step 6:

[0066] If the server detects a suspicious conversation, it automatically reports it to a fraud hotline, which includes the call details along with an automatically generated warning message.

[0067] Step 7:

[0068] The hotline receives the report and contacts the user. The hotline checks the situation and advises the user on the necessary measures (for example, freezing the account or contacting the police).

[0069] Step 8:

[0070] If the user wants to exclude certain callers from detection, the device provides an exclusion list management function. The user can add the IDs of specific callers (e.g., family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system.

[0071] Example 1

[0072] 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."

[0073] In recent years, the number of victims of telephone fraud has been increasing. Socially vulnerable people, such as the elderly, are often targeted, creating a need for an efficient system to prevent this from happening. However, conventional call monitoring systems lack the ability to detect suspicious conversations in real time, making it difficult to respond quickly. Furthermore, monitoring all calls poses the risk of privacy violations and interruption of important calls. Furthermore, systems lack the appropriate announcement function to enhance fraud deterrence.

[0074] 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.

[0075] In this invention, the server includes means for converting call content into text in real time, means for analyzing the converted call content using a generative model to detect suspicious conversations, means for automatically reporting to a reporting hotline when a suspicious conversation is detected, means for the terminal to play an announcement when a call starts, and means for managing a list of specific callers to be excluded from detection. This enables real-time monitoring and analysis of call content, and makes it possible to quickly report suspicious conversations while protecting privacy at the same time.

[0076] "Means of converting call content into text in real time" refers to technology that instantly converts voice data during a call into text information.

[0077] A "generative model" refers to an algorithm that uses machine learning to extract patterns from data and generate new information.

[0078] "Means for detecting suspicious conversations" refers to technology that analyzes the text content of phone calls and identifies conversations that may be fraudulent based on pre-set keywords and patterns.

[0079] "Means for automatically reporting to a reporting hotline" refers to a system that automatically sends the necessary information to a designated hotline when a suspicious conversation is detected.

[0080] "Means for the terminal to play an announcement when a call is initiated" refers to a function that automatically plays a pre-recorded message when a user initiates a call.

[0081] "Means for managing a list of specific callers to be excluded from detection" refers to a function for registering and managing callers that should not be monitored by the system.

[0082] The present invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model. Specifically, it detects suspicious conversations and automatically reports them to a hotline if necessary, thereby preventing fraud. The system configuration and operation are described in detail below.

[0083] System configuration

[0084] This system consists of the following main components:

[0085] 1. Terminal

[0086] The device through which a user makes a call, including smartphones and landlines.

[0087] The device is responsible for collecting voice data and sending it to the server.

[0088] It has the ability to play an announcement at the start of a call saying "This conversation is being recorded."

[0089] 2. Server

[0090] The received voice data is converted into text in real time.

[0091] The textual data is input into a generative AI model to detect suspicious conversations.

[0092] If a suspicious conversation is detected, a report will be automatically sent to the reporting hotline.

[0093] The generative AI model used is, for example, OpenAI's GPT-4.

[0094] 3. Reporting Desk

[0095] It has the function of receiving reports from the server, contacting the relevant user, and taking necessary measures.

[0096] Hardware and Software Examples

[0097] Speech recognition service: Converts voice data into text using services such as Google Cloud Speech-to-Text API and Amazon Transcribe.

[0098] Generative AI model: Analyzes the text of phone calls using models such as OpenAI's GPT-4.

[0099] Communication protocol: The HTTPS protocol is used to send audio data from the device to the server.

[0100] System operation explanation

[0101] 1. Start a call

[0102] When a user starts a call, the device automatically plays an announcement saying, "This conversation is being recorded."

[0103] 2. Collection and transmission of voice data

[0104] A microphone installed on the device collects voice data during the call and transmits it to the server in real time.

[0105] 3. Speech-to-text

[0106] The server uses a speech recognition service to convert the received voice data into text in real time.

[0107] 4. Text Data Analysis

[0108] The transcribed conversations are fed into a generative AI model that detects specific keywords and phrases to identify suspicious conversations.

[0109] For example, input the following prompt into the generative model:

[0110] Text of the call: "This is an urgent call from your bank. Please deposit money into XXX immediately."

[0111] Keywords to detect: deposit, bank, transfer, PIN, emergency

[0112] 5. Detecting and reporting suspicious conversations

[0113] When the generative model detects suspicious keywords such as "deposit" or "bank," the server automatically sends the relevant call details and a warning message to the reporting hotline.

[0114] 6. Responses via the reporting desk

[0115] The reporting hotline will contact the user in question, confirm the possibility of fraud, and take the necessary measures.

[0116] Specific examples

[0117] Consider a case where an elderly user receives a fraudulent phone call and is tricked into making a deposit, claiming to be an "urgent call from the bank." As the call begins, the device plays an announcement saying, "This conversation is being recorded," which is expected to have a deterrent effect on fraudulent activity. Collected voice data is converted into text in real time and analyzed by a generative model on the server. If the generative model detects suspicious keywords and determines that the conversation is suspicious, the server automatically notifies the hotline. The hotline then contacts the user and takes action to prevent fraud.

[0118] In this way, the present invention provides a system that effectively prevents fraud against elderly people and others by monitoring and analyzing call content in real time and immediately reporting any suspicious content.

[0119] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0120] Step 1: User initiates a call

[0121] A user initiates a call using a device. Specifically, the user launches a phone app, enters the other party's phone number, and presses the call button. This operation establishes the call.

[0122] Input: The phone number of the person you want to call

[0123] Output: Established call session

[0124] Step 2: The device plays an announcement

[0125] The device will play an announcement saying, "This conversation is being recorded." Specifically, the device will play an announcement message from its built-in speaker to let the other party know that the call is being recorded.

[0126] Input: Established call session

[0127] Output: Announcement message played

[0128] Step 3: The device collects audio data

[0129] The device collects voice data in real time during a call. Specifically, the device's microphone captures the voice of the call and stores the data in an internal buffer.

[0130] Input: Audio during a call

[0131] Output: Collected audio data

[0132] Step 4: The device sends the audio data to the server

[0133] The device sends the collected voice data to the server in real time. Specifically, the device uploads the voice data to the server using the HTTPS protocol.

[0134] Input: Collected audio data

[0135] Output: Audio data sent to the server

[0136] Step 5: The server converts the audio data into text

[0137] The server converts the received voice data into text in real time using a voice recognition service. Specifically, the server sends the voice data to a voice recognition API (e.g., Google Cloud Speech-to-Text API) and obtains the converted text.

[0138] Input: Received audio data

[0139] Output: Text of the call

[0140] Step 6: The server inputs the text data into the generative model

[0141] The server inputs the textual content of the call into a generative AI model, which analyzes it. Specifically, the server inputs the text data into a generative AI model (e.g., OpenAI's GPT-4) and generates prompts to detect suspicious keywords and phrases.

[0142] Input: Transcribed call

[0143] Output: Suspicious keywords and phrases detected

[0144] Step 7: The server detects suspicious conversations

[0145] The server detects suspicious conversations based on the output of the generative model. Specifically, the server analyzes the detection results from the generative model and determines whether specific keywords (e.g., "deposit," "bank," or "transfer") are included.

[0146] Input: Detection result

[0147] Output: Suspicious conversation detection flag

[0148] Step 8: The server automatically notifies the reporting point

[0149] If a suspicious conversation is detected, the server automatically reports it to the reporting desk. Specifically, the server generates the report content (textualized call content and a warning message) and sends a POST request to the reporting desk's API endpoint.

[0150] Input: Suspicious conversation detection flag, textual content of the call

[0151] Output: Report data to reporting desk

[0152] Step 9: The hotline contacts the user

[0153] The reporting desk will contact the relevant user, confirm the possibility of fraud, and take necessary measures. Specifically, the reporting desk operator will contact the user by phone or email to confirm the details of the situation.

[0154] Input: Report data

[0155] Output: User contact data

[0156] Step 10: Your device manages the exclusion list

[0157] To exclude specific callers from detection, the user manages an exclusion list on the device. Specifically, the user adds specific phone numbers to the exclusion list from the device's settings screen, and the device synchronizes the exclusion list with the server.

[0158] Input: Phone number of the person you want to add

[0159] Output: Updated exclusion list

[0160] (Application example 1)

[0161] 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."

[0162] While systems exist that monitor phone call content and detect suspicious conversations, they lack the ability to respond in real time or to quickly detect and report suspicious conversations in real-world environments. Furthermore, there is no mechanism for immediately notifying security personnel, making it difficult to prevent fraud and criminal activity. The present invention aims to solve these problems by providing a system that monitors and detects suspicious conversations in real time and quickly reports and notifies users.

[0163] 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.

[0164] In this invention, the server includes means for converting call content into text in real time, means for analyzing the converted call content using a generative model to detect suspicious conversations, means for automatically reporting to a fraud hotline when a suspicious conversation is detected, means for capturing the conversation content in real time using speech recognition and analyzing the suspicious conversation using a generative model, and means for sending an alert to a security officer when a suspicious conversation is detected. This makes it possible to monitor suspicious conversations in real time and respond quickly to prevent fraud and criminal acts.

[0165] "Means for converting call content into text in real time" refers to a function that converts voice data during a call into text format in real time, thereby recording the content of the call as text information.

[0166] "Means of analyzing textual content of phone calls using a generative model to detect suspicious conversations" refers to a function that uses a generative AI model to analyze textual content of phone calls and identify conversations that may involve fraudulent activity.

[0167] "Means of automatically reporting to a fraud hotline when a suspicious conversation is detected" refers to a function that automatically reports to a specialized agency responsible for eliminating fraud when a suspicious conversation is detected.

[0168] "Means of capturing conversation content in real time using voice recognition and analyzing suspicious conversations using a generative model" refers to a function that uses voice recognition technology to capture conversation content in real time, analyzes that data using a generative AI model, and identifies suspicious content.

[0169] "Means for sending alerts to security personnel when suspicious conversations are detected" refers to a function that quickly sends an alert to security personnel when suspicious conversations are discovered.

[0170] "A means of playing an announcement at the start of a call saying 'This conversation is being recorded'" refers to a function that plays a voice announcement at the start of a call to notify the other party that the conversation will be recorded.

[0171] "Means for managing a list that excludes conversations of specific callers from detection targets" refers to a function that excludes conversations of callers specified by the user from monitoring targets and manages that list.

[0172] A "reporting hotline" refers to a specialized agency or department that receives contact when suspicious conversations are detected and takes appropriate action.

[0173] A "generative model" refers to an AI algorithm or machine learning model that analyzes input data and generates new information or results, such as GPT-3 or BERT.

[0174] "Sending an alert" refers to the action of sending a warning or notification to a pre-defined notification destination when a problem occurs.

[0175] The present invention relates to a system that monitors telephone conversations in real time and quickly detects and reports suspicious conversations. The system configuration and processing flow for implementing the present invention will be described in detail below.

[0176] System Configuration

[0177] This system is broadly composed of the following three components:

[0178] 1. User terminal: refers to devices such as smart glasses, smartphones, head-mounted displays, etc. These devices have the ability to capture voice data in real time and transmit that data to a server.

[0179] 2. Server: This server has speech recognition capabilities to receive voice data and convert it to text, as well as the ability to analyze the text data using a generative AI model.

[0180] 3. Reporting system: This system sends an alert to a fraud reporting hotline or security officer if the server detects any suspicious conversations.

[0181] Program processing

[0182] User terminal

[0183] The user device captures the audio at the beginning of the call and plays an announcement saying, "This conversation is being recorded." The audio data is transmitted to the server in real time. The user device can perform this function using, for example, smart glasses or a smartphone.

[0184] server

[0185] The server receives the voice data sent from the user's device and converts it into text using speech recognition technology. Commonly used technologies on the server include Google Speech-to-Text API and IBM Watson.

[0186] The transcribed data is then fed into a generative AI model (e.g., GPT-3 or BERT), which has algorithms that identify specific keywords and patterns in the text to detect suspicious conversations.

[0187] Example of an input prompt for a generative AI model:

[0188] "Please analyze the following conversation. If there are any suspicious or dangerous statements, please mark them as suspicious. Conversation: 'We are going to give this confidential information to our business partner.'"

[0189] Reporting System

[0190] If a suspicious conversation is detected, the server automatically sends an alert to a fraud hotline or security officer, including the text of the suspicious conversation and any necessary warning messages.

[0191] Specific examples

[0192] For example, imagine a person in a confidential company meeting says, "I'm going to give this confidential information to a third party." This conversation is captured by smart glasses, and the audio data is sent to a server. The audio is converted into text in real time and analyzed by a generative AI model. If the model detects suspicious keywords such as "confidential information" or "give to a third party," an alert is immediately sent to security personnel. This rapid reporting prevents fraudulent activity before it occurs.

[0193] Through the system configuration and processing described above, the present invention realizes early detection of suspicious conversations in real time and rapid response.

[0194] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0195] Step 1:

[0196] At the start of a call, the user terminal captures the audio data and plays an announcement saying, "This conversation is being recorded."

[0197] Input: Call start trigger

[0198] Output: Announcement audio, captured audio data

[0199] How it works: Your smart glasses or smartphone will start recording audio through the microphone and play back an announcement.

[0200] Step 2:

[0201] The captured audio data is sent to the server in real time.

[0202] Input: Captured audio data

[0203] Output: Audio data sent to the server

[0204] Operation: The user terminal uses the communication module to send voice data to the server.

[0205] Step 3:

[0206] The server converts the received voice data into text using voice recognition technology.

[0207] Input: Audio data received on the server

[0208] Output: Text of the call

[0209] How it works: The server converts the audio data into text using the Google Speech-to-Text API or IBM Watson.

[0210] Step 4:

[0211] Using a generative AI model, the text of phone calls is analyzed to detect suspicious conversations.

[0212] Input: Transcribed call transcript

[0213] Output: Suspicious conversation detection result (suspicious / normal)

[0214] How it works: The server feeds text data into a generative AI model (e.g., GPT-3 or BERT) to analyze it for relevant keywords and patterns.

[0215] Step 5:

[0216] If a suspicious conversation is detected, the server sends an alert to a fraud hotline or security officer.

[0217] Input: Suspicious conversation detection results and corresponding text content

[0218] Output: Informational and warning messages

[0219] How it works: The server notifies security personnel via email or a dedicated application.

[0220] Step 6:

[0221] If the user wants to exclude a caller from detection, they can add and manage the ID of the specific caller to the exclusion list.

[0222] Input: User-specified callee ID

[0223] Output: Updated exclusion list

[0224] Operation: To exclude the user's specified caller from detection, the ID is sent to the server's exclusion list and the list is updated.

[0225] 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.

[0226] This invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model and an emotion engine. It detects suspicious conversations and automatically reports them to a fraud hotline as needed, preventing fraud before it happens. The system's program processing is explained in detail below.

[0227] System Overview

[0228] In this system, when a user initiates a call, the device collects voice data and the server converts the data into text in real time. In addition, the server analyzes the text of the call using a generative model and an emotion engine to detect suspicious conversations. If a suspicious conversation is detected, the server automatically reports it to a fraud hotline and the necessary measures are taken.

[0229] Announcement at the start of a call

[0230] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," which acts as a deterrent to fraudulent activity on the other end of the call.

[0231] Speech to text

[0232] During a call, voice data is collected by the user's device and sent to a server, which then uses a voice recognition service to convert the voice data into text in real time, thereby recording the contents of the call as text information.

[0233] Text Analysis

[0234] The server inputs the real-time text of the call into a generative model for analysis. The generative model uses natural language processing (NLP) algorithms to analyze the text and detect specific keywords and phrases. In addition, an emotion engine recognizes the user's emotions and incorporates that emotional information into the analysis results.

[0235] The role of the emotional engine

[0236] The emotion engine analyzes emotions from the user's tone of voice and facial expressions during a call and uses this information to determine suspicious conversations detected by the generative model. Specifically, if the emotion engine detects that the user is feeling strong anxiety or confusion, the generative model will take this emotional information into account and determine that there is a high possibility of fraud.

[0237] Automatic reporting of suspicious conversations

[0238] When the generative model and emotion engine work together to detect a suspicious conversation, the server automatically reports it to a fraud hotline. The report includes the content of the call, the generated warning message, and the user's emotional state. The hotline then contacts the user, confirms the situation, and takes necessary measures.

[0239] Managing the Exclusion List

[0240] Users can also exclude specific callers from detection. The device has an exclusion list management function that allows users to add the IDs of specific callers (for example, family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system, so the system can be operated without interfering with important calls.

[0241] Specific examples

[0242] For example, consider the case where an elderly user receives a fraudulent phone call, claiming to be an "urgent call from the bank" and requesting a deposit. When the user starts the call, the device announces, "This conversation is being recorded," and voice data is collected. This voice data is converted into text in real time and analyzed on the server using a generative model and emotion engine. If the generative model detects keywords such as "deposit" and "bank," and the emotion engine detects high anxiety as the user's emotional state, the server determines this to be a suspicious conversation and automatically reports it to a fraud hotline. The hotline contacts the user, confirms the situation, and instructs them to contact the bank or the police, for example, thereby preventing fraud before it occurs.

[0243] In this way, the present invention provides a system that can detect suspicious conversations in real time and respond quickly to them, effectively preventing fraud against elderly people and other victims. By combining it with an emotion engine, it becomes possible to detect suspicious conversations with even higher accuracy.

[0244] The processing flow will be explained below.

[0245] Step 1:

[0246] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," to deter the other party from engaging in fraudulent activity.

[0247] Step 2:

[0248] The device collects the audio data and sends the audio stream to the server. The audio stream is collected in real time to cover the entire call.

[0249] Step 3:

[0250] The server receives the audio stream and uses a speech recognition service to convert the audio data into text in real time, resulting in a written record of the conversation.

[0251] Step 4:

[0252] The server inputs the transcribed conversations into a generative model, which then uses natural language processing algorithms to analyze the text and detect specific keywords and phrases.

[0253] Step 5:

[0254] The server recognizes the user's emotions using an emotion engine, which analyzes the user's emotional state from the tone, rhythm, and speed of the voice.

[0255] Step 6:

[0256] The server combines the analysis results of the generative model and the emotion engine to determine whether the conversation is suspicious. For example, if keywords such as "deposit" and "PIN number" are detected and the emotion engine determines that the user is anxious or confused, it will determine that the conversation is suspicious.

[0257] Step 7:

[0258] If the generative model and emotion engine detect a suspicious conversation, the server automatically reports it to a fraud hotline, including the call content, a warning message, and the user's emotional state.

[0259] Step 8:

[0260] The hotline receives the report and contacts the user. The hotline checks the situation and advises the user on the necessary measures (for example, freezing the account or contacting the police).

[0261] Step 9:

[0262] If the user wants to exclude certain callers from detection, the device provides an exclusion list management function. The user can add the IDs of specific callers (e.g., family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system.

[0263] Step 10:

[0264] After the exclusion list management is applied, the device determines whether a new call is subject to detection based on the list. Calls with parties registered on the list are excluded from the analysis of the generative model and emotion engine and are treated as normal calls.

[0265] Example 2

[0266] 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."

[0267] In recent years, telephone fraud has been on the rise, with scams targeting the elderly in particular becoming more diverse. Under these circumstances, there is a need for a system that can monitor phone conversations in real time and quickly detect and report suspicious conversations. However, conventional systems face challenges in not only converting and analyzing phone conversations into text, but also in making highly accurate judgments that take into account the user's emotions. Another issue is the lack of functionality for flexibly managing the targets of monitoring so as not to interfere with routine or important calls.

[0268] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0269] In this invention, the server includes a means for converting call content into text in real time, a means for analyzing the converted call content using a generative model to detect suspicious conversations, a means for automatically reporting to a fraud hotline when a suspicious conversation is detected, and a means for analyzing the user's emotions using an emotion analysis engine and reflecting the results in determining whether the conversation is suspicious. This enables highly accurate detection and prompt reporting of suspicious conversations that take into account not only the call content but also the user's emotions. Furthermore, by including a means for announcing "This conversation is being recorded" at the start of a call and a means for managing a list of specific callers to be excluded from detection, it is possible to minimize interference with everyday calls while increasing the deterrent effect on the caller.

[0270] "Call content" refers to the voice information uttered by the user during a call.

[0271] "Real-time" is a time concept that refers to processing occurring almost simultaneously with the moment a call is made.

[0272] "Texting" is the process of converting audio information into written information.

[0273] A "generative model" is an algorithm that performs natural language processing and analyzes input text data to extract specific information.

[0274] "Suspicious conversations" refer to conversations that contain keywords or phrases that could be fraudulent.

[0275] An "emotion analysis engine" is software for analyzing a user's emotional state, detecting emotions from the tone and content of the voice during a call.

[0276] A "fraud reporting hotline" refers to an institution or service that accepts reports when suspicious conversations are detected.

[0277] An "announcement" is a voice message played from the terminal at the start of a call, informing the other party that the call will be recorded.

[0278] An "exclusion list" is a list in which the IDs of specific callers are registered so that calls with those people are not subject to monitoring.

[0279] The present invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model and an emotion analysis engine. It detects suspicious conversations and automatically reports them to a fraud hotline as needed, thereby preventing fraud damage. A specific embodiment of the system is described below.

[0280] System Overview

[0281] In this system, when a user initiates a call, the device collects voice data, and the server converts the voice data into text in real time. In addition, the server analyzes the text of the call using a generative model and an emotion analysis engine to detect suspicious conversations. If a suspicious conversation is detected, the server automatically reports it to a fraud hotline, and necessary measures are taken.

[0282] Announcement at the start of a call

[0283] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," which acts as a deterrent to fraudulent activity on the other end of the call.

[0284] Speech to text

[0285] During a call, the user's device collects and transmits voice data to a server, which then uses a speech recognition service (e.g., Google Cloud Speech-to-Text) to convert the data into text in real time, thereby recording the contents of the call as text information.

[0286] Text Analysis

[0287] The server inputs the real-time text of the call into a generative model (e.g., OpenAI GPT-4) for analysis. The generative model uses natural language processing (NLP) algorithms to analyze the text and detect specific keywords and phrases. In addition, a sentiment analysis engine (e.g., IBM Watson Tone Analyzer) recognizes the user's emotions and incorporates that emotional information into the analysis results.

[0288] The role of sentiment analysis engines

[0289] The emotion analysis engine analyzes emotions from the tone and content of the user's voice during a call and uses this information to determine suspicious conversations detected by the generative model. Specifically, if the emotion analysis engine detects that the user is feeling strong anxiety or confusion, the generative model will take this emotional information into account and determine that there is a high possibility of fraud.

[0290] Automatic reporting of suspicious conversations

[0291] When the generative model and sentiment analysis engine work together to detect a suspicious conversation, the server automatically reports it to a fraud hotline. The report includes the content of the call, the generated warning message, and the user's emotional state. The hotline then contacts the user to confirm the situation and take necessary measures.

[0292] Managing the Exclusion List

[0293] Users can also exclude specific callers from detection. The device has an exclusion list management function that allows users to add the IDs of specific callers (for example, family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system, so the system can be operated without interfering with important calls.

[0294] Specific examples

[0295] For example, consider the case where an elderly user receives a fraudulent phone call, claiming to be an "urgent call from the bank" and requesting a deposit. When the user starts the call, the device announces, "This conversation is being recorded," and voice data is collected. This voice data is converted into text in real time and analyzed on the server using a generative model and emotion analysis engine. If the generative model detects keywords such as "deposit" and "bank," and the emotion analysis engine detects high anxiety as the user's emotional state, the server determines this to be a suspicious conversation and automatically reports it to a fraud hotline. The hotline contacts the user, confirms the situation, and instructs them to contact the bank or the police, for example, thereby preventing the fraud from occurring.

[0296] Prompt Sentence Examples

[0297] "The user is on a call and the other party claims to be making an urgent call from their bank, requesting a deposit. The user's emotional state indicates high anxiety. Please determine whether this call is likely to be fraudulent."

[0298] The purpose of this system is to effectively prevent fraud, especially among the elderly, by detecting suspicious conversations in real time and responding quickly. By combining it with an emotion analysis engine, it is possible to detect suspicious conversations with even higher accuracy.

[0299] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0300] Step 1:

[0301] When a user starts a call, the device plays an announcement saying, "This conversation is being recorded." Specifically, the device's audio playback function is used to play a pre-recorded announcement. This announcement acts as a deterrent to the other party.

[0302] Step 2:

[0303] The terminal collects voice data uttered by the user during a call. This voice data is captured through the terminal's microphone and temporarily stored in buffer memory. The input is the user's voice, and the output is the voice data stored in the buffer memory.

[0304] Step 3:

[0305] The collected voice data is sent to the server in real time. Data is sent from the terminal to the server via network communication. The input is the voice data in the buffer memory, and the output is the voice data received on the server.

[0306] Step 4:

[0307] The server converts the received voice data into text using a voice recognition service. Specifically, it calls a voice recognition API such as Google Cloud Speech-to-Text and converts the voice data into text data. The input is the voice data, and the output is the text of the conversation.

[0308] Step 5:

[0309] The server inputs the converted conversation data into a generative model for analysis. The generative model uses natural language processing algorithms such as OpenAI GPT-4. The input is text data, and the output is the analysis results (e.g., detection of specific keywords or phrases).

[0310] Step 6:

[0311] At the same time, the server uses an emotion analysis engine to analyze the user's emotions. Specifically, it uses IBM Watson Tone Analyzer to extract emotions from the content of the call and the tone of the voice. The input is text data, and the output is emotional data.

[0312] Step 7:

[0313] The analysis results of the generative model are combined with the emotional data from the emotion analysis engine to identify suspicious conversations. The server then inputs this data into the algorithm to determine whether there is a possibility of fraud. The inputs are the analysis results and emotional data, and the output is the judgment result.

[0314] Step 8:

[0315] If a suspicious conversation is detected, the server automatically reports it to a fraud hotline. The report includes the text of the conversation, a generated warning message, and the user's emotional state. The input is the judgment result and related data, and the output is a report message.

[0316] Step 9:

[0317] Users manage an exclusion list to exclude specific callers from detection. Users can add the IDs of specific callers to the exclusion list through their devices. This information is sent from the device to the server. The input is the user's operation, and the output is the updated exclusion list.

[0318] (Application example 2)

[0319] 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."

[0320] In recent years, the number of fraudulent phone calls has been increasing, with elderly people and corporate employees being particularly targeted. When receiving a suspicious call, it is currently difficult to quickly and accurately assess the content and take appropriate measures. Furthermore, conventional methods for analyzing call content have low accuracy and there is a risk of suspicious conversations being overlooked, so an effective system to prevent fraudulent phone calls is needed.

[0321] 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.

[0322] In this invention, the server includes means for converting the contents of the call into text in real time, means for analyzing the converted contents of the call using a generative model to detect suspicious conversations, means for analyzing the emotional state during the call in real time using an emotion analysis engine, and means for automatically reporting to a fraud hotline if a suspicious conversation is detected. This makes it possible to analyze the contents of the call in real time, detect suspicious conversations with high accuracy, and take prompt measures.

[0323] "Means of converting phone conversations into text in real time" refers to a system that uses voice recognition technology to instantly convert conversations that take place during a phone call into text information.

[0324] "Means for detecting suspicious conversations through analysis using a generative model" is a system that analyzes the text content of conversations using a generative model based on a natural language processing algorithm to identify suspicious conversations that may involve fraud or dishonesty.

[0325] An "emotion analysis engine" refers to a program or device that evaluates emotions from the tone of voice and linguistic expressions during a call and analyzes that information in real time.

[0326] "Means for automatically reporting to fraud hotlines" refers to a system that automatically notifies fraud prevention officers and agencies when suspicious conversations are detected.

[0327] "Means for playing announcements" refers to a system that has the function of automatically playing a notification voice message such as "This conversation is being recorded" at the start of a call.

[0328] "Means for managing a list that excludes conversations of specific callers from detection" refers to a database or program that registers and manages information about pre-designated callers in order to exclude their conversations from monitoring.

[0329] The present invention is a system that converts telephone conversations into text in real time and analyzes the converted text using a generative model and an emotion analysis engine. The system detects suspicious conversations and automatically reports them to a fraud hotline, thereby preventing fraud damage. The specific configuration and operation of the system for implementing the present invention are described below.

[0330] System configuration

[0331] 1. Device:

[0332] The device has the ability to record the contents of calls and plays an announcement at the start of the call saying, "This conversation is being recorded." The recorded voice data is converted into text in real time and sent to a server.

[0333] 2. Server:

[0334] The server has the following main functions:

[0335] Speech recognition engine: Using the Python speech_recognition library, recorded voice data is converted into text in real time. This function allows the contents of calls to be instantly recorded as text information on the server.

[0336] Generative Model: Using Natural Language Processing (NLP) algorithms, Hugging Face's transformers library is used to analyze the transcribed conversations. This generative model detects specific keywords and phrases associated with fraud and scams.

[0337] Emotion analysis engine: Evaluates emotions from speech tone and linguistic expressions. This allows the user's emotional state to be analyzed in real time and reflected in the generative model's decisions.

[0338] Automatic reporting function: If a suspicious conversation is detected, a notification containing the call content and analysis results will be automatically sent to a fraud reporting hotline. This function uses the Python requests library.

[0339] Specific examples

[0340] For example, consider the case where an elderly user receives a fraudulent phone call requesting that the user deposit money into their bank account. When the user starts the call, the device plays a message saying, "This conversation is being recorded," and begins collecting voice data. This voice data is converted into text in real time and sent to a server.

[0341] The server uses a speech recognition engine to convert voice data into text and inputs the text content into a generative model for analysis. It also uses an emotion analysis engine to evaluate the user's emotional state, and if high anxiety is detected, it reflects this in the generative model's judgment. Specifically, it detects keywords such as "deposit" and "bank" from the text content, and if it determines that the user is in a very anxious state, the server recognizes this as a suspicious conversation.

[0342] The automated reporting function then automatically sends a notification containing the call details and analysis results to the fraud hotline, allowing the hotline to respond quickly and prevent damage before it occurs.

[0343] Prompt Sentence Examples

[0344] "Enter the text of the call below to analyze it for potential fraud."

[0345] Call: "Please deposit money into my bank account."

[0346] Emotional state: "Very anxious"

[0347] In this way, the system of the present invention provides an effective means of preventing fraud by detecting suspicious conversations in real time with high accuracy and responding quickly.By combining a speech recognition engine, a generative model, and a sentiment analysis engine, we can build a very powerful security service.

[0348] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0349] Step 1:

[0350] When a user starts a call, the device plays an announcement saying "This conversation is being recorded." This serves as a warning to the other party and a deterrent to fraudulent activity. The input is the call start event, and the output is the playback of the announcement audio.

[0351] Step 2:

[0352] The terminal starts collecting voice data in real time during the call. The voice data is obtained from the microphone and stored in a buffer. The input is the continuous voice data from the microphone, and the output is the collected voice data buffer.

[0353] Step 3:

[0354] The collected voice data is sent from the device to the server. The input is the data buffer of the device, and the output is the voice data transferred to the server. Specifically, the data is sent using the device's communication module.

[0355] Step 4:

[0356] The server converts the received voice data into text in real time using a speech recognition engine. A speech recognition engine (e.g., Python's speech_recognition library) is used to generate text data from the input voice data. The input is voice data, and the output is the text of the conversation.

[0357] Step 5:

[0358] The server inputs the converted conversation into a generative model (e.g., Hugging Face's transformers library) and analyzes it using natural language processing algorithms to detect specific keywords and phrases related to fraud. The input is text data, and the output is the results of keyword detection.

[0359] Step 6:

[0360] The server uses an emotion analysis engine to evaluate the text and emotional state in real time. The emotion analysis engine evaluates emotions from speech tone and linguistic expressions and incorporates this information into the analysis results of the generative model. The input is audio data and text data, and the output is the detected emotional state.

[0361] Step 7:

[0362] The server integrates the results of the generative model and the sentiment analysis engine to determine whether suspicious conversations have been detected. The inputs are the keyword detection results from the generative model and the evaluation results from the sentiment analysis engine, and the output is the detection results of suspicious conversations.

[0363] Step 8:

[0364] If a suspicious conversation is detected, the server automatically sends a notification to the fraud hotline, including the call content and analysis results. The notification also includes the user's emotional state. The input is the detection result of the suspicious conversation, and the output is a notification message to the fraud hotline. Specifically, an HTTP request is sent using the Python requests library.

[0365] Step 9:

[0366] The server manages a list of users to exclude conversations with specific callers from detection. It stores the information of the callers specified by the user in a database and manages detection targets based on that list. The input is the list of callers specified by the user, and the output is the updated exclusion list.

[0367] This series of processes enables the system to quickly and accurately detect suspicious conversations and prevent fraud before it occurs.

[0368] 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.

[0369] 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.

[0370] 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.

[0371] [Second embodiment]

[0372] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0373] 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.

[0374] 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).

[0375] 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.

[0376] 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.

[0377] 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).

[0378] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0379] 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.

[0380] 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.

[0381] 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.

[0382] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0383] 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."

[0384] The present invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model. Specifically, it detects suspicious conversations and automatically reports them to a fraud hotline as necessary, thereby preventing fraud damage. The system's program processing is described in detail below.

[0385] System Overview

[0386] In this system, when a user initiates a call, the device collects voice data, and the server converts the voice data into text in real time. In addition, the server uses a generative model to analyze the text of the call and detect suspicious conversations. If a suspicious conversation is detected, the server automatically reports it to a fraud hotline, and the necessary measures are taken.

[0387] Announcement at the start of a call

[0388] When a user starts a call, the device plays an announcement saying, "This conversation is being recorded." This announcement acts as a deterrent to fraudulent activity on the other end of the call.

[0389] Speech to text

[0390] During a call, voice data is collected by the user's device and sent to a server, which then uses a voice recognition service to convert the voice data into text in real time, thereby recording the contents of the call as text information.

[0391] Text Analysis

[0392] The server inputs the real-time text of the call into a generative model for analysis. The generative model has an algorithm that detects specific keywords and phrases and identifies suspicious conversations. For example, if keywords such as "deposit," "PIN number," or "transfer" are detected, it will determine that there is a possibility of fraud.

[0393] Automatic reporting of suspicious conversations

[0394] If the generative model detects a suspicious conversation, the server automatically reports it to a fraud hotline. The report includes the content of the call and the generated warning message. The hotline then contacts the user, confirms the situation, and takes any necessary measures.

[0395] Managing the Exclusion List

[0396] Users can also exclude specific callers from detection. The device has an exclusion list management function that allows users to add the IDs of specific callers (for example, family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system, so the system can be operated without interfering with important calls.

[0397] Specific examples

[0398] Consider a case where an elderly user receives a fraudulent phone call, claiming to be an "urgent call from the bank," and is tricked into making a deposit. When the call begins, the device announces, "This conversation is being recorded," and audio data is collected. This audio data is converted into text in real time and analyzed by a generative model on a server. If the generative model detects keywords such as "deposit" and "bank" and determines that the conversation is suspicious, the server automatically reports the incident to a fraud hotline. The hotline contacts the user, confirms the situation, and takes necessary measures to prevent fraud before it occurs.

[0399] In this way, the present invention provides a system that can detect suspicious conversations in real time and respond quickly, thereby effectively preventing fraud against elderly people and others.

[0400] The processing flow will be explained below.

[0401] Step 1:

[0402] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," to deter the other party from engaging in fraudulent activity.

[0403] Step 2:

[0404] The device collects the audio data and sends the audio stream to the server. The audio stream is collected in real time to cover the entire call.

[0405] Step 3:

[0406] The server receives the audio stream and uses a speech recognition service to convert the audio data into text in real time, resulting in a written record of the conversation.

[0407] Step 4:

[0408] The server inputs the transcribed conversations into a generative model, which uses natural language processing (NLP) algorithms to analyze the text and detect specific keywords and phrases.

[0409] Step 5:

[0410] The generative model detects suspicious conversations. For example, if a conversation contains keywords such as "deposit," "PIN number," or "transfer," it determines that it may be fraudulent.

[0411] Step 6:

[0412] If the server detects a suspicious conversation, it automatically reports it to a fraud hotline, which includes the call details along with an automatically generated warning message.

[0413] Step 7:

[0414] The hotline receives the report and contacts the user. The hotline checks the situation and advises the user on the necessary measures (for example, freezing the account or contacting the police).

[0415] Step 8:

[0416] If the user wants to exclude certain callers from detection, the device provides an exclusion list management function. The user can add the IDs of specific callers (e.g., family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system.

[0417] Example 1

[0418] 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."

[0419] In recent years, the number of victims of telephone fraud has been increasing. Socially vulnerable people, such as the elderly, are often targeted, creating a need for an efficient system to prevent this from happening. However, conventional call monitoring systems lack the ability to detect suspicious conversations in real time, making it difficult to respond quickly. Furthermore, monitoring all calls poses the risk of privacy violations and interruption of important calls. Furthermore, systems lack the appropriate announcement function to enhance fraud deterrence.

[0420] 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.

[0421] In this invention, the server includes means for converting call content into text in real time, means for analyzing the converted call content using a generative model to detect suspicious conversations, means for automatically reporting to a reporting hotline when a suspicious conversation is detected, means for the terminal to play an announcement when a call starts, and means for managing a list of specific callers to be excluded from detection. This enables real-time monitoring and analysis of call content, and makes it possible to quickly report suspicious conversations while protecting privacy at the same time.

[0422] "Means of converting call content into text in real time" refers to technology that instantly converts voice data during a call into text information.

[0423] A "generative model" refers to an algorithm that uses machine learning to extract patterns from data and generate new information.

[0424] "Means for detecting suspicious conversations" refers to technology that analyzes the text content of phone calls and identifies conversations that may be fraudulent based on pre-set keywords and patterns.

[0425] "Means for automatically reporting to a reporting hotline" refers to a system that automatically sends the necessary information to a designated hotline when a suspicious conversation is detected.

[0426] "Means for the terminal to play an announcement when a call is initiated" refers to a function that automatically plays a pre-recorded message when a user initiates a call.

[0427] "Means for managing a list of specific callers to be excluded from detection" refers to a function for registering and managing callers that should not be monitored by the system.

[0428] The present invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model. Specifically, it detects suspicious conversations and automatically reports them to a hotline if necessary, thereby preventing fraud. The system configuration and operation are described in detail below.

[0429] System configuration

[0430] This system consists of the following main components:

[0431] 1. Terminal

[0432] The device through which a user makes a call, including smartphones and landlines.

[0433] The device is responsible for collecting voice data and sending it to the server.

[0434] It has the ability to play an announcement at the start of a call saying "This conversation is being recorded."

[0435] 2. Server

[0436] The received voice data is converted into text in real time.

[0437] The textual data is input into a generative AI model to detect suspicious conversations.

[0438] If a suspicious conversation is detected, a report will be automatically sent to the reporting hotline.

[0439] The generative AI model used is, for example, OpenAI's GPT-4.

[0440] 3. Reporting Desk

[0441] It has the function of receiving reports from the server, contacting the relevant user, and taking necessary measures.

[0442] Hardware and Software Examples

[0443] Speech recognition service: Converts voice data into text using services such as Google Cloud Speech-to-Text API and Amazon Transcribe.

[0444] Generative AI model: Analyzes the text of phone calls using models such as OpenAI's GPT-4.

[0445] Communication protocol: The HTTPS protocol is used to send audio data from the device to the server.

[0446] System operation explanation

[0447] 1. Start a call

[0448] When a user starts a call, the device automatically plays an announcement saying, "This conversation is being recorded."

[0449] 2. Collection and transmission of voice data

[0450] A microphone installed on the device collects voice data during the call and transmits it to the server in real time.

[0451] 3. Speech-to-text

[0452] The server uses a speech recognition service to convert the received voice data into text in real time.

[0453] 4. Text Data Analysis

[0454] The transcribed conversations are fed into a generative AI model that detects specific keywords and phrases to identify suspicious conversations.

[0455] For example, input the following prompt into the generative model:

[0456] Text of the call: "This is an urgent call from your bank. Please deposit money into XXX immediately."

[0457] Keywords to detect: deposit, bank, transfer, PIN, emergency

[0458] 5. Detecting and reporting suspicious conversations

[0459] When the generative model detects suspicious keywords such as "deposit" or "bank," the server automatically sends the relevant call details and a warning message to the reporting hotline.

[0460] 6. Responses via the reporting desk

[0461] The reporting hotline will contact the user in question, confirm the possibility of fraud, and take the necessary measures.

[0462] Specific examples

[0463] Consider a case where an elderly user receives a fraudulent phone call and is tricked into making a deposit, claiming to be an "urgent call from the bank." As the call begins, the device plays an announcement saying, "This conversation is being recorded," which is expected to have a deterrent effect on fraudulent activity. Collected voice data is converted into text in real time and analyzed by a generative model on the server. If the generative model detects suspicious keywords and determines that the conversation is suspicious, the server automatically notifies the hotline. The hotline then contacts the user and takes action to prevent fraud.

[0464] In this way, the present invention provides a system that effectively prevents fraud against elderly people and others by monitoring and analyzing call content in real time and immediately reporting any suspicious content.

[0465] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0466] Step 1: User initiates a call

[0467] A user initiates a call using a device. Specifically, the user launches a phone app, enters the other party's phone number, and presses the call button. This operation establishes the call.

[0468] Input: The phone number of the person you want to call

[0469] Output: Established call session

[0470] Step 2: The device plays an announcement

[0471] The device will play an announcement saying, "This conversation is being recorded." Specifically, the device will play an announcement message from its built-in speaker to let the other party know that the call is being recorded.

[0472] Input: Established call session

[0473] Output: Announcement message played

[0474] Step 3: The device collects audio data

[0475] The device collects voice data in real time during a call. Specifically, the device's microphone captures the voice of the call and stores the data in an internal buffer.

[0476] Input: Audio during a call

[0477] Output: Collected audio data

[0478] Step 4: The device sends the audio data to the server

[0479] The device sends the collected voice data to the server in real time. Specifically, the device uploads the voice data to the server using the HTTPS protocol.

[0480] Input: Collected audio data

[0481] Output: Audio data sent to the server

[0482] Step 5: The server converts the audio data into text

[0483] The server converts the received voice data into text in real time using a voice recognition service. Specifically, the server sends the voice data to a voice recognition API (e.g., Google Cloud Speech-to-Text API) and obtains the converted text.

[0484] Input: Received audio data

[0485] Output: Text of the call

[0486] Step 6: The server inputs the text data into the generative model

[0487] The server inputs the textual content of the call into a generative AI model, which analyzes it. Specifically, the server inputs the text data into a generative AI model (e.g., OpenAI's GPT-4) and generates prompts to detect suspicious keywords and phrases.

[0488] Input: Transcribed call

[0489] Output: Suspicious keywords and phrases detected

[0490] Step 7: The server detects suspicious conversations

[0491] The server detects suspicious conversations based on the output of the generative model. Specifically, the server analyzes the detection results from the generative model and determines whether specific keywords (e.g., "deposit," "bank," or "transfer") are included.

[0492] Input: Detection result

[0493] Output: Suspicious conversation detection flag

[0494] Step 8: The server automatically notifies the reporting point

[0495] If a suspicious conversation is detected, the server automatically reports it to the reporting desk. Specifically, the server generates the report content (textualized call content and a warning message) and sends a POST request to the reporting desk's API endpoint.

[0496] Input: Suspicious conversation detection flag, textual content of the call

[0497] Output: Report data to reporting desk

[0498] Step 9: The hotline contacts the user

[0499] The reporting desk will contact the relevant user, confirm the possibility of fraud, and take necessary measures. Specifically, the reporting desk operator will contact the user by phone or email to confirm the details of the situation.

[0500] Input: Report data

[0501] Output: User contact data

[0502] Step 10: Your device manages the exclusion list

[0503] To exclude specific callers from detection, the user manages an exclusion list on the device. Specifically, the user adds specific phone numbers to the exclusion list from the device's settings screen, and the device synchronizes the exclusion list with the server.

[0504] Input: Phone number of the person you want to add

[0505] Output: Updated exclusion list

[0506] (Application example 1)

[0507] 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."

[0508] While systems exist that monitor phone call content and detect suspicious conversations, they lack the ability to respond in real time or to quickly detect and report suspicious conversations in real-world environments. Furthermore, there is no mechanism for immediately notifying security personnel, making it difficult to prevent fraud and criminal activity. The present invention aims to solve these problems by providing a system that monitors and detects suspicious conversations in real time and quickly reports and notifies users.

[0509] 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.

[0510] In this invention, the server includes means for converting call content into text in real time, means for analyzing the converted call content using a generative model to detect suspicious conversations, means for automatically reporting to a fraud hotline when a suspicious conversation is detected, means for capturing the conversation content in real time using speech recognition and analyzing the suspicious conversation using a generative model, and means for sending an alert to a security officer when a suspicious conversation is detected. This makes it possible to monitor suspicious conversations in real time and respond quickly to prevent fraud and criminal acts.

[0511] "Means for converting call content into text in real time" refers to a function that converts voice data during a call into text format in real time, thereby recording the content of the call as text information.

[0512] "Means of analyzing textual content of phone calls using a generative model to detect suspicious conversations" refers to a function that uses a generative AI model to analyze textual content of phone calls and identify conversations that may involve fraudulent activity.

[0513] "Means of automatically reporting to a fraud hotline when a suspicious conversation is detected" refers to a function that automatically reports to a specialized agency responsible for eliminating fraud when a suspicious conversation is detected.

[0514] "Means of capturing conversation content in real time using voice recognition and analyzing suspicious conversations using a generative model" refers to a function that uses voice recognition technology to capture conversation content in real time, analyzes that data using a generative AI model, and identifies suspicious content.

[0515] "Means for sending alerts to security personnel when suspicious conversations are detected" refers to a function that quickly sends an alert to security personnel when suspicious conversations are discovered.

[0516] "A means of playing an announcement at the start of a call saying 'This conversation is being recorded'" refers to a function that plays a voice announcement at the start of a call to notify the other party that the conversation will be recorded.

[0517] "Means for managing a list that excludes conversations of specific callers from detection targets" refers to a function that excludes conversations of callers specified by the user from monitoring targets and manages that list.

[0518] A "reporting hotline" refers to a specialized agency or department that receives contact when suspicious conversations are detected and takes appropriate action.

[0519] A "generative model" refers to an AI algorithm or machine learning model that analyzes input data and generates new information or results, such as GPT-3 or BERT.

[0520] "Sending an alert" refers to the action of sending a warning or notification to a pre-defined notification destination when a problem occurs.

[0521] The present invention relates to a system that monitors telephone conversations in real time and quickly detects and reports suspicious conversations. The system configuration and processing flow for implementing the present invention will be described in detail below.

[0522] System Configuration

[0523] This system is broadly composed of the following three components:

[0524] 1. User terminal: refers to devices such as smart glasses, smartphones, head-mounted displays, etc. These devices have the ability to capture voice data in real time and transmit that data to a server.

[0525] 2. Server: This server has speech recognition capabilities to receive voice data and convert it to text, as well as the ability to analyze the text data using a generative AI model.

[0526] 3. Reporting system: This system sends an alert to a fraud reporting hotline or security officer if the server detects any suspicious conversations.

[0527] Program processing

[0528] User terminal

[0529] The user device captures the audio at the beginning of the call and plays an announcement saying, "This conversation is being recorded." The audio data is transmitted to the server in real time. The user device can perform this function using, for example, smart glasses or a smartphone.

[0530] server

[0531] The server receives the voice data sent from the user's device and converts it into text using speech recognition technology. Commonly used technologies on the server include Google Speech-to-Text API and IBM Watson.

[0532] The transcribed data is then fed into a generative AI model (e.g., GPT-3 or BERT), which has algorithms that identify specific keywords and patterns in the text to detect suspicious conversations.

[0533] Example of an input prompt for a generative AI model:

[0534] "Please analyze the following conversation. If there are any suspicious or dangerous statements, please mark them as suspicious. Conversation: 'We are going to give this confidential information to our business partner.'"

[0535] Reporting System

[0536] If a suspicious conversation is detected, the server automatically sends an alert to a fraud hotline or security officer, including the text of the suspicious conversation and any necessary warning messages.

[0537] Specific examples

[0538] For example, imagine a person in a confidential company meeting says, "I'm going to give this confidential information to a third party." This conversation is captured by smart glasses, and the audio data is sent to a server. The audio is converted into text in real time and analyzed by a generative AI model. If the model detects suspicious keywords such as "confidential information" or "give to a third party," an alert is immediately sent to security personnel. This rapid reporting prevents fraudulent activity before it occurs.

[0539] Through the system configuration and processing described above, the present invention realizes early detection of suspicious conversations in real time and rapid response.

[0540] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0541] Step 1:

[0542] At the start of a call, the user terminal captures the audio data and plays an announcement saying, "This conversation is being recorded."

[0543] Input: Call start trigger

[0544] Output: Announcement audio, captured audio data

[0545] How it works: Your smart glasses or smartphone will start recording audio through the microphone and play back an announcement.

[0546] Step 2:

[0547] The captured audio data is sent to the server in real time.

[0548] Input: Captured audio data

[0549] Output: Audio data sent to the server

[0550] Operation: The user terminal uses the communication module to send voice data to the server.

[0551] Step 3:

[0552] The server converts the received voice data into text using voice recognition technology.

[0553] Input: Audio data received on the server

[0554] Output: Text of the call

[0555] How it works: The server converts the audio data into text using the Google Speech-to-Text API or IBM Watson.

[0556] Step 4:

[0557] Using a generative AI model, the text of phone calls is analyzed to detect suspicious conversations.

[0558] Input: Transcribed call transcript

[0559] Output: Suspicious conversation detection result (suspicious / normal)

[0560] How it works: The server feeds text data into a generative AI model (e.g., GPT-3 or BERT) to analyze it for relevant keywords and patterns.

[0561] Step 5:

[0562] If a suspicious conversation is detected, the server sends an alert to a fraud hotline or security officer.

[0563] Input: Suspicious conversation detection results and corresponding text content

[0564] Output: Informational and warning messages

[0565] How it works: The server notifies security personnel via email or a dedicated application.

[0566] Step 6:

[0567] If the user wants to exclude a caller from detection, they can add and manage the ID of the specific caller to the exclusion list.

[0568] Input: User-specified callee ID

[0569] Output: Updated exclusion list

[0570] Operation: To exclude the user's specified caller from detection, the ID is sent to the server's exclusion list and the list is updated.

[0571] 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.

[0572] This invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model and an emotion engine. It detects suspicious conversations and automatically reports them to a fraud hotline as needed, preventing fraud before it happens. The system's program processing is explained in detail below.

[0573] System Overview

[0574] In this system, when a user initiates a call, the device collects voice data and the server converts the data into text in real time. In addition, the server analyzes the text of the call using a generative model and an emotion engine to detect suspicious conversations. If a suspicious conversation is detected, the server automatically reports it to a fraud hotline and the necessary measures are taken.

[0575] Announcement at the start of a call

[0576] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," which acts as a deterrent to fraudulent activity on the other end of the call.

[0577] Speech to text

[0578] During a call, voice data is collected by the user's device and sent to a server, which then uses a voice recognition service to convert the voice data into text in real time, thereby recording the contents of the call as text information.

[0579] Text Analysis

[0580] The server inputs the real-time text of the call into a generative model for analysis. The generative model uses natural language processing (NLP) algorithms to analyze the text and detect specific keywords and phrases. In addition, an emotion engine recognizes the user's emotions and incorporates that emotional information into the analysis results.

[0581] The role of the emotional engine

[0582] The emotion engine analyzes emotions from the user's tone of voice and facial expressions during a call and uses this information to determine suspicious conversations detected by the generative model. Specifically, if the emotion engine detects that the user is feeling strong anxiety or confusion, the generative model will take this emotional information into account and determine that there is a high possibility of fraud.

[0583] Automatic reporting of suspicious conversations

[0584] When the generative model and emotion engine work together to detect a suspicious conversation, the server automatically reports it to a fraud hotline. The report includes the content of the call, the generated warning message, and the user's emotional state. The hotline then contacts the user, confirms the situation, and takes necessary measures.

[0585] Managing the Exclusion List

[0586] Users can also exclude specific callers from detection. The device has an exclusion list management function that allows users to add the IDs of specific callers (for example, family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system, so the system can be operated without interfering with important calls.

[0587] Specific examples

[0588] For example, consider the case where an elderly user receives a fraudulent phone call, claiming to be an "urgent call from the bank" and requesting a deposit. When the user starts the call, the device announces, "This conversation is being recorded," and voice data is collected. This voice data is converted into text in real time and analyzed on the server using a generative model and emotion engine. If the generative model detects keywords such as "deposit" and "bank," and the emotion engine detects high anxiety as the user's emotional state, the server determines this to be a suspicious conversation and automatically reports it to a fraud hotline. The hotline contacts the user, confirms the situation, and instructs them to contact the bank or the police, for example, thereby preventing fraud before it occurs.

[0589] In this way, the present invention provides a system that can detect suspicious conversations in real time and respond quickly to them, effectively preventing fraud against elderly people and other victims. By combining it with an emotion engine, it becomes possible to detect suspicious conversations with even higher accuracy.

[0590] The processing flow will be explained below.

[0591] Step 1:

[0592] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," to deter the other party from engaging in fraudulent activity.

[0593] Step 2:

[0594] The device collects the audio data and sends the audio stream to the server. The audio stream is collected in real time to cover the entire call.

[0595] Step 3:

[0596] The server receives the audio stream and uses a speech recognition service to convert the audio data into text in real time, resulting in a written record of the conversation.

[0597] Step 4:

[0598] The server inputs the transcribed conversations into a generative model, which then uses natural language processing algorithms to analyze the text and detect specific keywords and phrases.

[0599] Step 5:

[0600] The server recognizes the user's emotions using an emotion engine, which analyzes the user's emotional state from the tone, rhythm, and speed of the voice.

[0601] Step 6:

[0602] The server combines the analysis results of the generative model and the emotion engine to determine whether the conversation is suspicious. For example, if keywords such as "deposit" and "PIN number" are detected and the emotion engine determines that the user is anxious or confused, it will determine that the conversation is suspicious.

[0603] Step 7:

[0604] If the generative model and emotion engine detect a suspicious conversation, the server automatically reports it to a fraud hotline, including the call content, a warning message, and the user's emotional state.

[0605] Step 8:

[0606] The hotline receives the report and contacts the user. The hotline checks the situation and advises the user on the necessary measures (for example, freezing the account or contacting the police).

[0607] Step 9:

[0608] If the user wants to exclude certain callers from detection, the device provides an exclusion list management function. The user can add the IDs of specific callers (e.g., family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system.

[0609] Step 10:

[0610] After the exclusion list management is applied, the device determines whether a new call is subject to detection based on the list. Calls with parties registered on the list are excluded from the analysis of the generative model and emotion engine and are treated as normal calls.

[0611] Example 2

[0612] 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."

[0613] In recent years, telephone fraud has been on the rise, with scams targeting the elderly in particular becoming more diverse. Under these circumstances, there is a need for a system that can monitor phone conversations in real time and quickly detect and report suspicious conversations. However, conventional systems face challenges in not only converting and analyzing phone conversations into text, but also in making highly accurate judgments that take into account the user's emotions. Another issue is the lack of functionality for flexibly managing the targets of monitoring so as not to interfere with routine or important calls.

[0614] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0615] In this invention, the server includes a means for converting call content into text in real time, a means for analyzing the converted call content using a generative model to detect suspicious conversations, a means for automatically reporting to a fraud hotline when a suspicious conversation is detected, and a means for analyzing the user's emotions using an emotion analysis engine and reflecting the results in determining whether the conversation is suspicious. This enables highly accurate detection and prompt reporting of suspicious conversations that take into account not only the call content but also the user's emotions. Furthermore, by including a means for announcing "This conversation is being recorded" at the start of a call and a means for managing a list of specific callers to be excluded from detection, it is possible to minimize interference with everyday calls while increasing the deterrent effect on the caller.

[0616] "Call content" refers to the voice information uttered by the user during a call.

[0617] "Real-time" is a time concept that refers to processing occurring almost simultaneously with the moment a call is made.

[0618] "Texting" is the process of converting audio information into written information.

[0619] A "generative model" is an algorithm that performs natural language processing and analyzes input text data to extract specific information.

[0620] "Suspicious conversations" refer to conversations that contain keywords or phrases that could be fraudulent.

[0621] An "emotion analysis engine" is software for analyzing a user's emotional state, detecting emotions from the tone and content of the voice during a call.

[0622] A "fraud reporting hotline" refers to an institution or service that accepts reports when suspicious conversations are detected.

[0623] An "announcement" is a voice message played from the terminal at the start of a call, informing the other party that the call will be recorded.

[0624] An "exclusion list" is a list in which the IDs of specific callers are registered so that calls with those people are not subject to monitoring.

[0625] The present invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model and an emotion analysis engine. It detects suspicious conversations and automatically reports them to a fraud hotline as needed, thereby preventing fraud damage. A specific embodiment of the system is described below.

[0626] System Overview

[0627] In this system, when a user initiates a call, the device collects voice data, and the server converts the voice data into text in real time. In addition, the server analyzes the text of the call using a generative model and an emotion analysis engine to detect suspicious conversations. If a suspicious conversation is detected, the server automatically reports it to a fraud hotline, and necessary measures are taken.

[0628] Announcement at the start of a call

[0629] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," which acts as a deterrent to fraudulent activity on the other end of the call.

[0630] Speech to text

[0631] During a call, the user's device collects and transmits voice data to a server, which then uses a speech recognition service (e.g., Google Cloud Speech-to-Text) to convert the data into text in real time, thereby recording the contents of the call as text information.

[0632] Text Analysis

[0633] The server inputs the real-time text of the call into a generative model (e.g., OpenAI GPT-4) for analysis. The generative model uses natural language processing (NLP) algorithms to analyze the text and detect specific keywords and phrases. In addition, a sentiment analysis engine (e.g., IBM Watson Tone Analyzer) recognizes the user's emotions and incorporates that emotional information into the analysis results.

[0634] The role of sentiment analysis engines

[0635] The emotion analysis engine analyzes emotions from the tone and content of the user's voice during a call and uses this information to determine suspicious conversations detected by the generative model. Specifically, if the emotion analysis engine detects that the user is feeling strong anxiety or confusion, the generative model will take this emotional information into account and determine that there is a high possibility of fraud.

[0636] Automatic reporting of suspicious conversations

[0637] When the generative model and sentiment analysis engine work together to detect a suspicious conversation, the server automatically reports it to a fraud hotline. The report includes the content of the call, the generated warning message, and the user's emotional state. The hotline then contacts the user to confirm the situation and take necessary measures.

[0638] Managing the Exclusion List

[0639] Users can also exclude specific callers from detection. The device has an exclusion list management function that allows users to add the IDs of specific callers (for example, family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system, so the system can be operated without interfering with important calls.

[0640] Specific examples

[0641] For example, consider the case where an elderly user receives a fraudulent phone call, claiming to be an "urgent call from the bank" and requesting a deposit. When the user starts the call, the device announces, "This conversation is being recorded," and voice data is collected. This voice data is converted into text in real time and analyzed on the server using a generative model and emotion analysis engine. If the generative model detects keywords such as "deposit" and "bank," and the emotion analysis engine detects high anxiety as the user's emotional state, the server determines this to be a suspicious conversation and automatically reports it to a fraud hotline. The hotline contacts the user, confirms the situation, and instructs them to contact the bank or the police, for example, thereby preventing the fraud from occurring.

[0642] Prompt Sentence Examples

[0643] "The user is on a call and the other party claims to be making an urgent call from their bank, requesting a deposit. The user's emotional state indicates high anxiety. Please determine whether this call is likely to be fraudulent."

[0644] The purpose of this system is to effectively prevent fraud, especially among the elderly, by detecting suspicious conversations in real time and responding quickly. By combining it with an emotion analysis engine, it is possible to detect suspicious conversations with even higher accuracy.

[0645] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0646] Step 1:

[0647] When a user starts a call, the device plays an announcement saying, "This conversation is being recorded." Specifically, the device's audio playback function is used to play a pre-recorded announcement. This announcement acts as a deterrent to the other party.

[0648] Step 2:

[0649] The terminal collects voice data uttered by the user during a call. This voice data is captured through the terminal's microphone and temporarily stored in buffer memory. The input is the user's voice, and the output is the voice data stored in the buffer memory.

[0650] Step 3:

[0651] The collected voice data is sent to the server in real time. Data is sent from the terminal to the server via network communication. The input is the voice data in the buffer memory, and the output is the voice data received on the server.

[0652] Step 4:

[0653] The server converts the received voice data into text using a voice recognition service. Specifically, it calls a voice recognition API such as Google Cloud Speech-to-Text and converts the voice data into text data. The input is the voice data, and the output is the text of the conversation.

[0654] Step 5:

[0655] The server inputs the converted conversation data into a generative model for analysis. The generative model uses natural language processing algorithms such as OpenAI GPT-4. The input is text data, and the output is the analysis results (e.g., detection of specific keywords or phrases).

[0656] Step 6:

[0657] At the same time, the server uses an emotion analysis engine to analyze the user's emotions. Specifically, it uses IBM Watson Tone Analyzer to extract emotions from the content of the call and the tone of the voice. The input is text data, and the output is emotional data.

[0658] Step 7:

[0659] The analysis results of the generative model are combined with the emotional data from the emotion analysis engine to identify suspicious conversations. The server then inputs this data into the algorithm to determine whether there is a possibility of fraud. The inputs are the analysis results and emotional data, and the output is the judgment result.

[0660] Step 8:

[0661] If a suspicious conversation is detected, the server automatically reports it to a fraud hotline. The report includes the text of the conversation, a generated warning message, and the user's emotional state. The input is the judgment result and related data, and the output is a report message.

[0662] Step 9:

[0663] Users manage an exclusion list to exclude specific callers from detection. Users can add the IDs of specific callers to the exclusion list through their devices. This information is sent from the device to the server. The input is the user's operation, and the output is the updated exclusion list.

[0664] (Application example 2)

[0665] 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."

[0666] In recent years, the number of fraudulent phone calls has been increasing, with elderly people and corporate employees being particularly targeted. When receiving a suspicious call, it is currently difficult to quickly and accurately assess the content and take appropriate measures. Furthermore, conventional methods for analyzing call content have low accuracy and there is a risk of suspicious conversations being overlooked, so an effective system to prevent fraudulent phone calls is needed.

[0667] 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.

[0668] In this invention, the server includes means for converting the contents of the call into text in real time, means for analyzing the converted contents of the call using a generative model to detect suspicious conversations, means for analyzing the emotional state during the call in real time using an emotion analysis engine, and means for automatically reporting to a fraud hotline if a suspicious conversation is detected. This makes it possible to analyze the contents of the call in real time, detect suspicious conversations with high accuracy, and take prompt measures.

[0669] "Means of converting phone conversations into text in real time" refers to a system that uses voice recognition technology to instantly convert conversations that take place during a phone call into text information.

[0670] "Means for detecting suspicious conversations through analysis using a generative model" is a system that analyzes the text content of conversations using a generative model based on a natural language processing algorithm to identify suspicious conversations that may involve fraud or dishonesty.

[0671] An "emotion analysis engine" refers to a program or device that evaluates emotions from the tone of voice and linguistic expressions during a call and analyzes that information in real time.

[0672] "Means for automatically reporting to fraud hotlines" refers to a system that automatically notifies fraud prevention officers and agencies when suspicious conversations are detected.

[0673] "Means for playing announcements" refers to a system that has the function of automatically playing a notification voice message such as "This conversation is being recorded" at the start of a call.

[0674] "Means for managing a list that excludes conversations of specific callers from detection" refers to a database or program that registers and manages information about pre-designated callers in order to exclude their conversations from monitoring.

[0675] The present invention is a system that converts telephone conversations into text in real time and analyzes the converted text using a generative model and an emotion analysis engine. The system detects suspicious conversations and automatically reports them to a fraud hotline, thereby preventing fraud damage. The specific configuration and operation of the system for implementing the present invention are described below.

[0676] System configuration

[0677] 1. Device:

[0678] The device has the ability to record the contents of calls and plays an announcement at the start of the call saying, "This conversation is being recorded." The recorded voice data is converted into text in real time and sent to a server.

[0679] 2. Server:

[0680] The server has the following main functions:

[0681] Speech recognition engine: Using the Python speech_recognition library, recorded voice data is converted into text in real time. This function allows the contents of calls to be instantly recorded as text information on the server.

[0682] Generative Model: Using Natural Language Processing (NLP) algorithms, Hugging Face's transformers library is used to analyze the transcribed conversations. This generative model detects specific keywords and phrases associated with fraud and scams.

[0683] Emotion analysis engine: Evaluates emotions from speech tone and linguistic expressions. This allows the user's emotional state to be analyzed in real time and reflected in the generative model's decisions.

[0684] Automatic reporting function: If a suspicious conversation is detected, a notification containing the call content and analysis results will be automatically sent to a fraud reporting hotline. This function uses the Python requests library.

[0685] Specific examples

[0686] For example, consider the case where an elderly user receives a fraudulent phone call requesting that the user deposit money into their bank account. When the user starts the call, the device plays a message saying, "This conversation is being recorded," and begins collecting voice data. This voice data is converted into text in real time and sent to a server.

[0687] The server uses a speech recognition engine to convert voice data into text and inputs the text content into a generative model for analysis. It also uses an emotion analysis engine to evaluate the user's emotional state, and if high anxiety is detected, it reflects this in the generative model's judgment. Specifically, it detects keywords such as "deposit" and "bank" from the text content, and if it determines that the user is in a very anxious state, the server recognizes this as a suspicious conversation.

[0688] The automated reporting function then automatically sends a notification containing the call details and analysis results to the fraud hotline, allowing the hotline to respond quickly and prevent damage before it occurs.

[0689] Prompt Sentence Examples

[0690] "Enter the text of the call below to analyze it for potential fraud."

[0691] Call: "Please deposit money into my bank account."

[0692] Emotional state: "Very anxious"

[0693] In this way, the system of the present invention provides an effective means of preventing fraud by detecting suspicious conversations in real time with high accuracy and responding quickly.By combining a speech recognition engine, a generative model, and a sentiment analysis engine, we can build a very powerful security service.

[0694] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0695] Step 1:

[0696] When a user starts a call, the device plays an announcement saying "This conversation is being recorded." This serves as a warning to the other party and a deterrent to fraudulent activity. The input is the call start event, and the output is the playback of the announcement audio.

[0697] Step 2:

[0698] The terminal starts collecting voice data in real time during the call. The voice data is obtained from the microphone and stored in a buffer. The input is the continuous voice data from the microphone, and the output is the collected voice data buffer.

[0699] Step 3:

[0700] The collected voice data is sent from the device to the server. The input is the data buffer of the device, and the output is the voice data transferred to the server. Specifically, the data is sent using the device's communication module.

[0701] Step 4:

[0702] The server converts the received voice data into text in real time using a speech recognition engine. A speech recognition engine (e.g., Python's speech_recognition library) is used to generate text data from the input voice data. The input is voice data, and the output is the text of the conversation.

[0703] Step 5:

[0704] The server inputs the converted conversation into a generative model (e.g., Hugging Face's transformers library) and analyzes it using natural language processing algorithms to detect specific keywords and phrases related to fraud. The input is text data, and the output is the results of keyword detection.

[0705] Step 6:

[0706] The server uses an emotion analysis engine to evaluate the text and emotional state in real time. The emotion analysis engine evaluates emotions from speech tone and linguistic expressions and incorporates this information into the analysis results of the generative model. The input is audio data and text data, and the output is the detected emotional state.

[0707] Step 7:

[0708] The server integrates the results of the generative model and the sentiment analysis engine to determine whether suspicious conversations have been detected. The inputs are the keyword detection results from the generative model and the evaluation results from the sentiment analysis engine, and the output is the detection results of suspicious conversations.

[0709] Step 8:

[0710] If a suspicious conversation is detected, the server automatically sends a notification to the fraud hotline, including the call content and analysis results. The notification also includes the user's emotional state. The input is the detection result of the suspicious conversation, and the output is a notification message to the fraud hotline. Specifically, an HTTP request is sent using the Python requests library.

[0711] Step 9:

[0712] The server manages a list of users to exclude conversations with specific callers from detection. It stores the information of the callers specified by the user in a database and manages detection targets based on that list. The input is the list of callers specified by the user, and the output is the updated exclusion list.

[0713] This series of processes enables the system to quickly and accurately detect suspicious conversations and prevent fraud before it occurs.

[0714] 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.

[0715] 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.

[0716] 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.

[0717] [Third embodiment]

[0718] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0719] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0720] 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).

[0721] 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.

[0722] 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.

[0723] 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).

[0724] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0725] 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.

[0726] 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.

[0727] 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.

[0728] 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.

[0729] 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."

[0730] The present invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model. Specifically, it detects suspicious conversations and automatically reports them to a fraud hotline as necessary, thereby preventing fraud damage. The system's program processing is described in detail below.

[0731] System Overview

[0732] In this system, when a user initiates a call, the device collects voice data, and the server converts the voice data into text in real time. In addition, the server uses a generative model to analyze the text of the call and detect suspicious conversations. If a suspicious conversation is detected, the server automatically reports it to a fraud hotline, and the necessary measures are taken.

[0733] Announcement at the start of a call

[0734] When a user starts a call, the device plays an announcement saying, "This conversation is being recorded." This announcement acts as a deterrent to fraudulent activity on the other end of the call.

[0735] Speech to text

[0736] During a call, voice data is collected by the user's device and sent to a server, which then uses a voice recognition service to convert the voice data into text in real time, thereby recording the contents of the call as text information.

[0737] Text Analysis

[0738] The server inputs the real-time text of the call into a generative model for analysis. The generative model has an algorithm that detects specific keywords and phrases and identifies suspicious conversations. For example, if keywords such as "deposit," "PIN number," or "transfer" are detected, it will determine that there is a possibility of fraud.

[0739] Automatic reporting of suspicious conversations

[0740] If the generative model detects a suspicious conversation, the server automatically reports it to a fraud hotline. The report includes the content of the call and the generated warning message. The hotline then contacts the user, confirms the situation, and takes any necessary measures.

[0741] Managing the Exclusion List

[0742] Users can also exclude specific callers from detection. The device has an exclusion list management function that allows users to add the IDs of specific callers (for example, family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system, so the system can be operated without interfering with important calls.

[0743] Specific examples

[0744] Consider a case where an elderly user receives a fraudulent phone call, claiming to be an "urgent call from the bank," and is tricked into making a deposit. When the call begins, the device announces, "This conversation is being recorded," and audio data is collected. This audio data is converted into text in real time and analyzed by a generative model on a server. If the generative model detects keywords such as "deposit" and "bank" and determines that the conversation is suspicious, the server automatically reports the incident to a fraud hotline. The hotline contacts the user, confirms the situation, and takes necessary measures to prevent fraud before it occurs.

[0745] In this way, the present invention provides a system that can detect suspicious conversations in real time and respond quickly, thereby effectively preventing fraud against elderly people and others.

[0746] The processing flow will be explained below.

[0747] Step 1:

[0748] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," to deter the other party from engaging in fraudulent activity.

[0749] Step 2:

[0750] The device collects the audio data and sends the audio stream to the server. The audio stream is collected in real time to cover the entire call.

[0751] Step 3:

[0752] The server receives the audio stream and uses a speech recognition service to convert the audio data into text in real time, resulting in a written record of the conversation.

[0753] Step 4:

[0754] The server inputs the transcribed conversations into a generative model, which uses natural language processing (NLP) algorithms to analyze the text and detect specific keywords and phrases.

[0755] Step 5:

[0756] The generative model detects suspicious conversations. For example, if a conversation contains keywords such as "deposit," "PIN number," or "transfer," it determines that it may be fraudulent.

[0757] Step 6:

[0758] If the server detects a suspicious conversation, it automatically reports it to a fraud hotline, which includes the call details along with an automatically generated warning message.

[0759] Step 7:

[0760] The hotline receives the report and contacts the user. The hotline checks the situation and advises the user on the necessary measures (for example, freezing the account or contacting the police).

[0761] Step 8:

[0762] If the user wants to exclude certain callers from detection, the device provides an exclusion list management function. The user can add the IDs of specific callers (e.g., family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system.

[0763] Example 1

[0764] 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."

[0765] In recent years, the number of victims of telephone fraud has been increasing. Socially vulnerable people, such as the elderly, are often targeted, creating a need for an efficient system to prevent this from happening. However, conventional call monitoring systems lack the ability to detect suspicious conversations in real time, making it difficult to respond quickly. Furthermore, monitoring all calls poses the risk of privacy violations and interruption of important calls. Furthermore, systems lack the appropriate announcement function to enhance fraud deterrence.

[0766] 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.

[0767] In this invention, the server includes means for converting call content into text in real time, means for analyzing the converted call content using a generative model to detect suspicious conversations, means for automatically reporting to a reporting hotline when a suspicious conversation is detected, means for the terminal to play an announcement when a call starts, and means for managing a list of specific callers to be excluded from detection. This enables real-time monitoring and analysis of call content, and makes it possible to quickly report suspicious conversations while protecting privacy at the same time.

[0768] "Means of converting call content into text in real time" refers to technology that instantly converts voice data during a call into text information.

[0769] A "generative model" refers to an algorithm that uses machine learning to extract patterns from data and generate new information.

[0770] "Means for detecting suspicious conversations" refers to technology that analyzes the text content of phone calls and identifies conversations that may be fraudulent based on pre-set keywords and patterns.

[0771] "Means for automatically reporting to a reporting hotline" refers to a system that automatically sends the necessary information to a designated hotline when a suspicious conversation is detected.

[0772] "Means for the terminal to play an announcement when a call is initiated" refers to a function that automatically plays a pre-recorded message when a user initiates a call.

[0773] "Means for managing a list of specific callers to be excluded from detection" refers to a function for registering and managing callers that should not be monitored by the system.

[0774] The present invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model. Specifically, it detects suspicious conversations and automatically reports them to a hotline if necessary, thereby preventing fraud. The system configuration and operation are described in detail below.

[0775] System configuration

[0776] This system consists of the following main components:

[0777] 1. Terminal

[0778] The device through which a user makes a call, including smartphones and landlines.

[0779] The device is responsible for collecting voice data and sending it to the server.

[0780] It has the ability to play an announcement at the start of a call saying "This conversation is being recorded."

[0781] 2. Server

[0782] The received voice data is converted into text in real time.

[0783] The textual data is input into a generative AI model to detect suspicious conversations.

[0784] If a suspicious conversation is detected, a report will be automatically sent to the reporting hotline.

[0785] The generative AI model used is, for example, OpenAI's GPT-4.

[0786] 3. Reporting Desk

[0787] It has the function of receiving reports from the server, contacting the relevant user, and taking necessary measures.

[0788] Hardware and Software Examples

[0789] Speech recognition service: Converts voice data into text using services such as Google Cloud Speech-to-Text API and Amazon Transcribe.

[0790] Generative AI model: Analyzes the text of phone calls using models such as OpenAI's GPT-4.

[0791] Communication protocol: The HTTPS protocol is used to send audio data from the device to the server.

[0792] System operation explanation

[0793] 1. Start a call

[0794] When a user starts a call, the device automatically plays an announcement saying, "This conversation is being recorded."

[0795] 2. Collection and transmission of voice data

[0796] A microphone installed on the device collects voice data during the call and transmits it to the server in real time.

[0797] 3. Speech-to-text

[0798] The server uses a speech recognition service to convert the received voice data into text in real time.

[0799] 4. Text Data Analysis

[0800] The transcribed conversations are fed into a generative AI model that detects specific keywords and phrases to identify suspicious conversations.

[0801] For example, input the following prompt into the generative model:

[0802] Text of the call: "This is an urgent call from your bank. Please deposit money into XXX immediately."

[0803] Keywords to detect: deposit, bank, transfer, PIN, emergency

[0804] 5. Detecting and reporting suspicious conversations

[0805] When the generative model detects suspicious keywords such as "deposit" or "bank," the server automatically sends the relevant call details and a warning message to the reporting hotline.

[0806] 6. Responses via the reporting desk

[0807] The reporting hotline will contact the user in question, confirm the possibility of fraud, and take the necessary measures.

[0808] Specific examples

[0809] Consider a case where an elderly user receives a fraudulent phone call and is tricked into making a deposit, claiming to be an "urgent call from the bank." As the call begins, the device plays an announcement saying, "This conversation is being recorded," which is expected to have a deterrent effect on fraudulent activity. Collected voice data is converted into text in real time and analyzed by a generative model on the server. If the generative model detects suspicious keywords and determines that the conversation is suspicious, the server automatically notifies the hotline. The hotline then contacts the user and takes action to prevent fraud.

[0810] In this way, the present invention provides a system that effectively prevents fraud against elderly people and others by monitoring and analyzing call content in real time and immediately reporting any suspicious content.

[0811] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0812] Step 1: User initiates a call

[0813] A user initiates a call using a device. Specifically, the user launches a phone app, enters the other party's phone number, and presses the call button. This operation establishes the call.

[0814] Input: The phone number of the person you want to call

[0815] Output: Established call session

[0816] Step 2: The device plays an announcement

[0817] The device will play an announcement saying, "This conversation is being recorded." Specifically, the device will play an announcement message from its built-in speaker to let the other party know that the call is being recorded.

[0818] Input: Established call session

[0819] Output: Announcement message played

[0820] Step 3: The device collects audio data

[0821] The device collects voice data in real time during a call. Specifically, the device's microphone captures the voice of the call and stores the data in an internal buffer.

[0822] Input: Audio during a call

[0823] Output: Collected audio data

[0824] Step 4: The device sends the audio data to the server

[0825] The device sends the collected voice data to the server in real time. Specifically, the device uploads the voice data to the server using the HTTPS protocol.

[0826] Input: Collected audio data

[0827] Output: Audio data sent to the server

[0828] Step 5: The server converts the audio data into text

[0829] The server converts the received voice data into text in real time using a voice recognition service. Specifically, the server sends the voice data to a voice recognition API (e.g., Google Cloud Speech-to-Text API) and obtains the converted text.

[0830] Input: Received audio data

[0831] Output: Text of the call

[0832] Step 6: The server inputs the text data into the generative model

[0833] The server inputs the textual content of the call into a generative AI model, which analyzes it. Specifically, the server inputs the text data into a generative AI model (e.g., OpenAI's GPT-4) and generates prompts to detect suspicious keywords and phrases.

[0834] Input: Transcribed call

[0835] Output: Suspicious keywords and phrases detected

[0836] Step 7: The server detects suspicious conversations

[0837] The server detects suspicious conversations based on the output of the generative model. Specifically, the server analyzes the detection results from the generative model and determines whether specific keywords (e.g., "deposit," "bank," or "transfer") are included.

[0838] Input: Detection result

[0839] Output: Suspicious conversation detection flag

[0840] Step 8: The server automatically notifies the reporting point

[0841] If a suspicious conversation is detected, the server automatically reports it to the reporting desk. Specifically, the server generates the report content (textualized call content and a warning message) and sends a POST request to the reporting desk's API endpoint.

[0842] Input: Suspicious conversation detection flag, textual content of the call

[0843] Output: Report data to reporting desk

[0844] Step 9: The hotline contacts the user

[0845] The reporting desk will contact the relevant user, confirm the possibility of fraud, and take necessary measures. Specifically, the reporting desk operator will contact the user by phone or email to confirm the details of the situation.

[0846] Input: Report data

[0847] Output: User contact data

[0848] Step 10: Your device manages the exclusion list

[0849] To exclude specific callers from detection, the user manages an exclusion list on the device. Specifically, the user adds specific phone numbers to the exclusion list from the device's settings screen, and the device synchronizes the exclusion list with the server.

[0850] Input: Phone number of the person you want to add

[0851] Output: Updated exclusion list

[0852] (Application example 1)

[0853] 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."

[0854] While systems exist that monitor phone call content and detect suspicious conversations, they lack the ability to respond in real time or to quickly detect and report suspicious conversations in real-world environments. Furthermore, there is no mechanism for immediately notifying security personnel, making it difficult to prevent fraud and criminal activity. The present invention aims to solve these problems by providing a system that monitors and detects suspicious conversations in real time and quickly reports and notifies users.

[0855] 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.

[0856] In this invention, the server includes means for converting call content into text in real time, means for analyzing the converted call content using a generative model to detect suspicious conversations, means for automatically reporting to a fraud hotline when a suspicious conversation is detected, means for capturing the conversation content in real time using speech recognition and analyzing the suspicious conversation using a generative model, and means for sending an alert to a security officer when a suspicious conversation is detected. This makes it possible to monitor suspicious conversations in real time and respond quickly to prevent fraud and criminal acts.

[0857] "Means for converting call content into text in real time" refers to a function that converts voice data during a call into text format in real time, thereby recording the content of the call as text information.

[0858] "Means of analyzing textual content of phone calls using a generative model to detect suspicious conversations" refers to a function that uses a generative AI model to analyze textual content of phone calls and identify conversations that may involve fraudulent activity.

[0859] "Means of automatically reporting to a fraud hotline when a suspicious conversation is detected" refers to a function that automatically reports to a specialized agency responsible for eliminating fraud when a suspicious conversation is detected.

[0860] "Means of capturing conversation content in real time using voice recognition and analyzing suspicious conversations using a generative model" refers to a function that uses voice recognition technology to capture conversation content in real time, analyzes that data using a generative AI model, and identifies suspicious content.

[0861] "Means for sending alerts to security personnel when suspicious conversations are detected" refers to a function that quickly sends an alert to security personnel when suspicious conversations are discovered.

[0862] "A means of playing an announcement at the start of a call saying 'This conversation is being recorded'" refers to a function that plays a voice announcement at the start of a call to notify the other party that the conversation will be recorded.

[0863] "Means for managing a list that excludes conversations of specific callers from detection targets" refers to a function that excludes conversations of callers specified by the user from monitoring targets and manages that list.

[0864] A "reporting hotline" refers to a specialized agency or department that receives contact when suspicious conversations are detected and takes appropriate action.

[0865] A "generative model" refers to an AI algorithm or machine learning model that analyzes input data and generates new information or results, such as GPT-3 or BERT.

[0866] "Sending an alert" refers to the action of sending a warning or notification to a pre-defined notification destination when a problem occurs.

[0867] The present invention relates to a system that monitors telephone conversations in real time and quickly detects and reports suspicious conversations. The system configuration and processing flow for implementing the present invention will be described in detail below.

[0868] System Configuration

[0869] This system is broadly composed of the following three components:

[0870] 1. User terminal: refers to devices such as smart glasses, smartphones, head-mounted displays, etc. These devices have the ability to capture voice data in real time and transmit that data to a server.

[0871] 2. Server: This server has speech recognition capabilities to receive voice data and convert it to text, as well as the ability to analyze the text data using a generative AI model.

[0872] 3. Reporting system: This system sends an alert to a fraud reporting hotline or security officer if the server detects any suspicious conversations.

[0873] Program processing

[0874] User terminal

[0875] The user device captures the audio at the beginning of the call and plays an announcement saying, "This conversation is being recorded." The audio data is transmitted to the server in real time. The user device can perform this function using, for example, smart glasses or a smartphone.

[0876] server

[0877] The server receives the voice data sent from the user's device and converts it into text using speech recognition technology. Commonly used technologies on the server include Google Speech-to-Text API and IBM Watson.

[0878] The transcribed data is then fed into a generative AI model (e.g., GPT-3 or BERT), which has algorithms that identify specific keywords and patterns in the text to detect suspicious conversations.

[0879] Example of an input prompt for a generative AI model:

[0880] "Please analyze the following conversation. If there are any suspicious or dangerous statements, please mark them as suspicious. Conversation: 'We are going to give this confidential information to our business partner.'"

[0881] Reporting System

[0882] If a suspicious conversation is detected, the server automatically sends an alert to a fraud hotline or security officer, including the text of the suspicious conversation and any necessary warning messages.

[0883] Specific examples

[0884] For example, imagine a person in a confidential company meeting says, "I'm going to give this confidential information to a third party." This conversation is captured by smart glasses, and the audio data is sent to a server. The audio is converted into text in real time and analyzed by a generative AI model. If the model detects suspicious keywords such as "confidential information" or "give to a third party," an alert is immediately sent to security personnel. This rapid reporting prevents fraudulent activity before it occurs.

[0885] Through the system configuration and processing described above, the present invention realizes early detection of suspicious conversations in real time and rapid response.

[0886] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0887] Step 1:

[0888] At the start of a call, the user terminal captures the audio data and plays an announcement saying, "This conversation is being recorded."

[0889] Input: Call start trigger

[0890] Output: Announcement audio, captured audio data

[0891] How it works: Your smart glasses or smartphone will start recording audio through the microphone and play back an announcement.

[0892] Step 2:

[0893] The captured audio data is sent to the server in real time.

[0894] Input: Captured audio data

[0895] Output: Audio data sent to the server

[0896] Operation: The user terminal uses the communication module to send voice data to the server.

[0897] Step 3:

[0898] The server converts the received voice data into text using voice recognition technology.

[0899] Input: Audio data received on the server

[0900] Output: Text of the call

[0901] How it works: The server converts the audio data into text using the Google Speech-to-Text API or IBM Watson.

[0902] Step 4:

[0903] Using a generative AI model, the text of phone calls is analyzed to detect suspicious conversations.

[0904] Input: Transcribed call transcript

[0905] Output: Suspicious conversation detection result (suspicious / normal)

[0906] How it works: The server feeds text data into a generative AI model (e.g., GPT-3 or BERT) to analyze it for relevant keywords and patterns.

[0907] Step 5:

[0908] If a suspicious conversation is detected, the server sends an alert to a fraud hotline or security officer.

[0909] Input: Suspicious conversation detection results and corresponding text content

[0910] Output: Informational and warning messages

[0911] How it works: The server notifies security personnel via email or a dedicated application.

[0912] Step 6:

[0913] If the user wants to exclude a caller from detection, they can add and manage the ID of the specific caller to the exclusion list.

[0914] Input: User-specified callee ID

[0915] Output: Updated exclusion list

[0916] Operation: To exclude the user's specified caller from detection, the ID is sent to the server's exclusion list and the list is updated.

[0917] 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.

[0918] This invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model and an emotion engine. It detects suspicious conversations and automatically reports them to a fraud hotline as needed, preventing fraud before it happens. The system's program processing is explained in detail below.

[0919] System Overview

[0920] In this system, when a user initiates a call, the device collects voice data and the server converts the data into text in real time. In addition, the server analyzes the text of the call using a generative model and an emotion engine to detect suspicious conversations. If a suspicious conversation is detected, the server automatically reports it to a fraud hotline and the necessary measures are taken.

[0921] Announcement at the start of a call

[0922] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," which acts as a deterrent to fraudulent activity on the other end of the call.

[0923] Speech to text

[0924] During a call, voice data is collected by the user's device and sent to a server, which then uses a voice recognition service to convert the voice data into text in real time, thereby recording the contents of the call as text information.

[0925] Text Analysis

[0926] The server inputs the real-time text of the call into a generative model for analysis. The generative model uses natural language processing (NLP) algorithms to analyze the text and detect specific keywords and phrases. In addition, an emotion engine recognizes the user's emotions and incorporates that emotional information into the analysis results.

[0927] The role of the emotional engine

[0928] The emotion engine analyzes emotions from the user's tone of voice and facial expressions during a call and uses this information to determine suspicious conversations detected by the generative model. Specifically, if the emotion engine detects that the user is feeling strong anxiety or confusion, the generative model will take this emotional information into account and determine that there is a high possibility of fraud.

[0929] Automatic reporting of suspicious conversations

[0930] When the generative model and emotion engine work together to detect a suspicious conversation, the server automatically reports it to a fraud hotline. The report includes the content of the call, the generated warning message, and the user's emotional state. The hotline then contacts the user, confirms the situation, and takes necessary measures.

[0931] Managing the Exclusion List

[0932] Users can also exclude specific callers from detection. The device has an exclusion list management function that allows users to add the IDs of specific callers (for example, family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system, so the system can be operated without interfering with important calls.

[0933] Specific examples

[0934] For example, consider the case where an elderly user receives a fraudulent phone call, claiming to be an "urgent call from the bank" and requesting a deposit. When the user starts the call, the device announces, "This conversation is being recorded," and voice data is collected. This voice data is converted into text in real time and analyzed on the server using a generative model and emotion engine. If the generative model detects keywords such as "deposit" and "bank," and the emotion engine detects high anxiety as the user's emotional state, the server determines this to be a suspicious conversation and automatically reports it to a fraud hotline. The hotline contacts the user, confirms the situation, and instructs them to contact the bank or the police, for example, thereby preventing fraud before it occurs.

[0935] In this way, the present invention provides a system that can detect suspicious conversations in real time and respond quickly to them, effectively preventing fraud against elderly people and other victims. By combining it with an emotion engine, it becomes possible to detect suspicious conversations with even higher accuracy.

[0936] The processing flow will be explained below.

[0937] Step 1:

[0938] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," to deter the other party from engaging in fraudulent activity.

[0939] Step 2:

[0940] The device collects the audio data and sends the audio stream to the server. The audio stream is collected in real time to cover the entire call.

[0941] Step 3:

[0942] The server receives the audio stream and uses a speech recognition service to convert the audio data into text in real time, resulting in a written record of the conversation.

[0943] Step 4:

[0944] The server inputs the transcribed conversations into a generative model, which then uses natural language processing algorithms to analyze the text and detect specific keywords and phrases.

[0945] Step 5:

[0946] The server recognizes the user's emotions using an emotion engine, which analyzes the user's emotional state from the tone, rhythm, and speed of the voice.

[0947] Step 6:

[0948] The server combines the analysis results of the generative model and the emotion engine to determine whether the conversation is suspicious. For example, if keywords such as "deposit" and "PIN number" are detected and the emotion engine determines that the user is anxious or confused, it will determine that the conversation is suspicious.

[0949] Step 7:

[0950] If the generative model and emotion engine detect a suspicious conversation, the server automatically reports it to a fraud hotline, including the call content, a warning message, and the user's emotional state.

[0951] Step 8:

[0952] The hotline receives the report and contacts the user. The hotline checks the situation and advises the user on the necessary measures (for example, freezing the account or contacting the police).

[0953] Step 9:

[0954] If the user wants to exclude certain callers from detection, the device provides an exclusion list management function. The user can add the IDs of specific callers (e.g., family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system.

[0955] Step 10:

[0956] After the exclusion list management is applied, the device determines whether a new call is subject to detection based on the list. Calls with parties registered on the list are excluded from the analysis of the generative model and emotion engine and are treated as normal calls.

[0957] Example 2

[0958] 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."

[0959] In recent years, telephone fraud has been on the rise, with scams targeting the elderly in particular becoming more diverse. Under these circumstances, there is a need for a system that can monitor phone conversations in real time and quickly detect and report suspicious conversations. However, conventional systems face challenges in not only converting and analyzing phone conversations into text, but also in making highly accurate judgments that take into account the user's emotions. Another issue is the lack of functionality for flexibly managing the targets of monitoring so as not to interfere with routine or important calls.

[0960] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0961] In this invention, the server includes a means for converting call content into text in real time, a means for analyzing the converted call content using a generative model to detect suspicious conversations, a means for automatically reporting to a fraud hotline when a suspicious conversation is detected, and a means for analyzing the user's emotions using an emotion analysis engine and reflecting the results in determining whether the conversation is suspicious. This enables highly accurate detection and prompt reporting of suspicious conversations that take into account not only the call content but also the user's emotions. Furthermore, by including a means for announcing "This conversation is being recorded" at the start of a call and a means for managing a list of specific callers to be excluded from detection, it is possible to minimize interference with everyday calls while increasing the deterrent effect on the caller.

[0962] "Call content" refers to the voice information uttered by the user during a call.

[0963] "Real-time" is a time concept that refers to processing occurring almost simultaneously with the moment a call is made.

[0964] "Texting" is the process of converting audio information into written information.

[0965] A "generative model" is an algorithm that performs natural language processing and analyzes input text data to extract specific information.

[0966] "Suspicious conversations" refer to conversations that contain keywords or phrases that could be fraudulent.

[0967] An "emotion analysis engine" is software for analyzing a user's emotional state, detecting emotions from the tone and content of the voice during a call.

[0968] A "fraud reporting hotline" refers to an institution or service that accepts reports when suspicious conversations are detected.

[0969] An "announcement" is a voice message played from the terminal at the start of a call, informing the other party that the call will be recorded.

[0970] An "exclusion list" is a list in which the IDs of specific callers are registered so that calls with those people are not subject to monitoring.

[0971] The present invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model and an emotion analysis engine. It detects suspicious conversations and automatically reports them to a fraud hotline as needed, thereby preventing fraud damage. A specific embodiment of the system is described below.

[0972] System Overview

[0973] In this system, when a user initiates a call, the device collects voice data, and the server converts the voice data into text in real time. In addition, the server analyzes the text of the call using a generative model and an emotion analysis engine to detect suspicious conversations. If a suspicious conversation is detected, the server automatically reports it to a fraud hotline, and necessary measures are taken.

[0974] Announcement at the start of a call

[0975] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," which acts as a deterrent to fraudulent activity on the other end of the call.

[0976] Speech to text

[0977] During a call, the user's device collects and transmits voice data to a server, which then uses a speech recognition service (e.g., Google Cloud Speech-to-Text) to convert the data into text in real time, thereby recording the contents of the call as text information.

[0978] Text Analysis

[0979] The server inputs the real-time text of the call into a generative model (e.g., OpenAI GPT-4) for analysis. The generative model uses natural language processing (NLP) algorithms to analyze the text and detect specific keywords and phrases. In addition, a sentiment analysis engine (e.g., IBM Watson Tone Analyzer) recognizes the user's emotions and incorporates that emotional information into the analysis results.

[0980] The role of sentiment analysis engines

[0981] The emotion analysis engine analyzes emotions from the tone and content of the user's voice during a call and uses this information to determine suspicious conversations detected by the generative model. Specifically, if the emotion analysis engine detects that the user is feeling strong anxiety or confusion, the generative model will take this emotional information into account and determine that there is a high possibility of fraud.

[0982] Automatic reporting of suspicious conversations

[0983] When the generative model and sentiment analysis engine work together to detect a suspicious conversation, the server automatically reports it to a fraud hotline. The report includes the content of the call, the generated warning message, and the user's emotional state. The hotline then contacts the user to confirm the situation and take necessary measures.

[0984] Managing the Exclusion List

[0985] Users can also exclude specific callers from detection. The device has an exclusion list management function that allows users to add the IDs of specific callers (for example, family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system, so the system can be operated without interfering with important calls.

[0986] Specific examples

[0987] For example, consider the case where an elderly user receives a fraudulent phone call, claiming to be an "urgent call from the bank" and requesting a deposit. When the user starts the call, the device announces, "This conversation is being recorded," and voice data is collected. This voice data is converted into text in real time and analyzed on the server using a generative model and emotion analysis engine. If the generative model detects keywords such as "deposit" and "bank," and the emotion analysis engine detects high anxiety as the user's emotional state, the server determines this to be a suspicious conversation and automatically reports it to a fraud hotline. The hotline contacts the user, confirms the situation, and instructs them to contact the bank or the police, for example, thereby preventing the fraud from occurring.

[0988] Prompt Sentence Examples

[0989] "The user is on a call and the other party claims to be making an urgent call from their bank, requesting a deposit. The user's emotional state indicates high anxiety. Please determine whether this call is likely to be fraudulent."

[0990] The purpose of this system is to effectively prevent fraud, especially among the elderly, by detecting suspicious conversations in real time and responding quickly. By combining it with an emotion analysis engine, it is possible to detect suspicious conversations with even higher accuracy.

[0991] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0992] Step 1:

[0993] When a user starts a call, the device plays an announcement saying, "This conversation is being recorded." Specifically, the device's audio playback function is used to play a pre-recorded announcement. This announcement acts as a deterrent to the other party.

[0994] Step 2:

[0995] The terminal collects voice data uttered by the user during a call. This voice data is captured through the terminal's microphone and temporarily stored in buffer memory. The input is the user's voice, and the output is the voice data stored in the buffer memory.

[0996] Step 3:

[0997] The collected voice data is sent to the server in real time. Data is sent from the terminal to the server via network communication. The input is the voice data in the buffer memory, and the output is the voice data received on the server.

[0998] Step 4:

[0999] The server converts the received voice data into text using a voice recognition service. Specifically, it calls a voice recognition API such as Google Cloud Speech-to-Text and converts the voice data into text data. The input is the voice data, and the output is the text of the conversation.

[1000] Step 5:

[1001] The server inputs the converted conversation data into a generative model for analysis. The generative model uses natural language processing algorithms such as OpenAI GPT-4. The input is text data, and the output is the analysis results (e.g., detection of specific keywords or phrases).

[1002] Step 6:

[1003] At the same time, the server uses an emotion analysis engine to analyze the user's emotions. Specifically, it uses IBM Watson Tone Analyzer to extract emotions from the content of the call and the tone of the voice. The input is text data, and the output is emotional data.

[1004] Step 7:

[1005] The analysis results of the generative model are combined with the emotional data from the emotion analysis engine to identify suspicious conversations. The server then inputs this data into the algorithm to determine whether there is a possibility of fraud. The inputs are the analysis results and emotional data, and the output is the judgment result.

[1006] Step 8:

[1007] If a suspicious conversation is detected, the server automatically reports it to a fraud hotline. The report includes the text of the conversation, a generated warning message, and the user's emotional state. The input is the judgment result and related data, and the output is a report message.

[1008] Step 9:

[1009] Users manage an exclusion list to exclude specific callers from detection. Users can add the IDs of specific callers to the exclusion list through their devices. This information is sent from the device to the server. The input is the user's operation, and the output is the updated exclusion list.

[1010] (Application example 2)

[1011] 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."

[1012] In recent years, the number of fraudulent phone calls has been increasing, with elderly people and corporate employees being particularly targeted. When receiving a suspicious call, it is currently difficult to quickly and accurately assess the content and take appropriate measures. Furthermore, conventional methods for analyzing call content have low accuracy and there is a risk of suspicious conversations being overlooked, so an effective system to prevent fraudulent phone calls is needed.

[1013] 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.

[1014] In this invention, the server includes means for converting the contents of the call into text in real time, means for analyzing the converted contents of the call using a generative model to detect suspicious conversations, means for analyzing the emotional state during the call in real time using an emotion analysis engine, and means for automatically reporting to a fraud hotline if a suspicious conversation is detected. This makes it possible to analyze the contents of the call in real time, detect suspicious conversations with high accuracy, and take prompt measures.

[1015] "Means of converting phone conversations into text in real time" refers to a system that uses voice recognition technology to instantly convert conversations that take place during a phone call into text information.

[1016] "Means for detecting suspicious conversations through analysis using a generative model" is a system that analyzes the text content of conversations using a generative model based on a natural language processing algorithm to identify suspicious conversations that may involve fraud or dishonesty.

[1017] An "emotion analysis engine" refers to a program or device that evaluates emotions from the tone of voice and linguistic expressions during a call and analyzes that information in real time.

[1018] "Means for automatically reporting to fraud hotlines" refers to a system that automatically notifies fraud prevention officers and agencies when suspicious conversations are detected.

[1019] "Means for playing announcements" refers to a system that has the function of automatically playing a notification voice message such as "This conversation is being recorded" at the start of a call.

[1020] "Means for managing a list that excludes conversations of specific callers from detection" refers to a database or program that registers and manages information about pre-designated callers in order to exclude their conversations from monitoring.

[1021] The present invention is a system that converts telephone conversations into text in real time and analyzes the converted text using a generative model and an emotion analysis engine. The system detects suspicious conversations and automatically reports them to a fraud hotline, thereby preventing fraud damage. The specific configuration and operation of the system for implementing the present invention are described below.

[1022] System configuration

[1023] 1. Device:

[1024] The device has the ability to record the contents of calls and plays an announcement at the start of the call saying, "This conversation is being recorded." The recorded voice data is converted into text in real time and sent to a server.

[1025] 2. Server:

[1026] The server has the following main functions:

[1027] Speech recognition engine: Using the Python speech_recognition library, recorded voice data is converted into text in real time. This function allows the contents of calls to be instantly recorded as text information on the server.

[1028] Generative Model: Using Natural Language Processing (NLP) algorithms, Hugging Face's transformers library is used to analyze the transcribed conversations. This generative model detects specific keywords and phrases associated with fraud and scams.

[1029] Emotion analysis engine: Evaluates emotions from speech tone and linguistic expressions. This allows the user's emotional state to be analyzed in real time and reflected in the generative model's decisions.

[1030] Automatic reporting function: If a suspicious conversation is detected, a notification containing the call content and analysis results will be automatically sent to a fraud reporting hotline. This function uses the Python requests library.

[1031] Specific examples

[1032] For example, consider the case where an elderly user receives a fraudulent phone call requesting that the user deposit money into their bank account. When the user starts the call, the device plays a message saying, "This conversation is being recorded," and begins collecting voice data. This voice data is converted into text in real time and sent to a server.

[1033] The server uses a speech recognition engine to convert voice data into text and inputs the text content into a generative model for analysis. It also uses an emotion analysis engine to evaluate the user's emotional state, and if high anxiety is detected, it reflects this in the generative model's judgment. Specifically, it detects keywords such as "deposit" and "bank" from the text content, and if it determines that the user is in a very anxious state, the server recognizes this as a suspicious conversation.

[1034] The automated reporting function then automatically sends a notification containing the call details and analysis results to the fraud hotline, allowing the hotline to respond quickly and prevent damage before it occurs.

[1035] Prompt Sentence Examples

[1036] "Enter the text of the call below to analyze it for potential fraud."

[1037] Call: "Please deposit money into my bank account."

[1038] Emotional state: "Very anxious"

[1039] In this way, the system of the present invention provides an effective means of preventing fraud by detecting suspicious conversations in real time with high accuracy and responding quickly.By combining a speech recognition engine, a generative model, and a sentiment analysis engine, we can build a very powerful security service.

[1040] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1041] Step 1:

[1042] When a user starts a call, the device plays an announcement saying "This conversation is being recorded." This serves as a warning to the other party and a deterrent to fraudulent activity. The input is the call start event, and the output is the playback of the announcement audio.

[1043] Step 2:

[1044] The terminal starts collecting voice data in real time during the call. The voice data is obtained from the microphone and stored in a buffer. The input is the continuous voice data from the microphone, and the output is the collected voice data buffer.

[1045] Step 3:

[1046] The collected voice data is sent from the device to the server. The input is the data buffer of the device, and the output is the voice data transferred to the server. Specifically, the data is sent using the device's communication module.

[1047] Step 4:

[1048] The server converts the received voice data into text in real time using a speech recognition engine. A speech recognition engine (e.g., Python's speech_recognition library) is used to generate text data from the input voice data. The input is voice data, and the output is the text of the conversation.

[1049] Step 5:

[1050] The server inputs the converted conversation into a generative model (e.g., Hugging Face's transformers library) and analyzes it using natural language processing algorithms to detect specific keywords and phrases related to fraud. The input is text data, and the output is the results of keyword detection.

[1051] Step 6:

[1052] The server uses an emotion analysis engine to evaluate the text and emotional state in real time. The emotion analysis engine evaluates emotions from speech tone and linguistic expressions and incorporates this information into the analysis results of the generative model. The input is audio data and text data, and the output is the detected emotional state.

[1053] Step 7:

[1054] The server integrates the results of the generative model and the sentiment analysis engine to determine whether suspicious conversations have been detected. The inputs are the keyword detection results from the generative model and the evaluation results from the sentiment analysis engine, and the output is the detection results of suspicious conversations.

[1055] Step 8:

[1056] If a suspicious conversation is detected, the server automatically sends a notification to the fraud hotline, including the call content and analysis results. The notification also includes the user's emotional state. The input is the detection result of the suspicious conversation, and the output is a notification message to the fraud hotline. Specifically, an HTTP request is sent using the Python requests library.

[1057] Step 9:

[1058] The server manages a list of users to exclude conversations with specific callers from detection. It stores the information of the callers specified by the user in a database and manages detection targets based on that list. The input is the list of callers specified by the user, and the output is the updated exclusion list.

[1059] This series of processes enables the system to quickly and accurately detect suspicious conversations and prevent fraud before it occurs.

[1060] 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.

[1061] 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.

[1062] 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.

[1063] [Fourth embodiment]

[1064] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1065] 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.

[1066] 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).

[1067] 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.

[1068] 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.

[1069] 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).

[1070] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1071] 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.

[1072] 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.

[1073] 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.

[1074] 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.

[1075] 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.

[1076] 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."

[1077] The present invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model. Specifically, it detects suspicious conversations and automatically reports them to a fraud hotline as necessary, thereby preventing fraud damage. The system's program processing is described in detail below.

[1078] System Overview

[1079] In this system, when a user initiates a call, the device collects voice data, and the server converts the voice data into text in real time. In addition, the server uses a generative model to analyze the text of the call and detect suspicious conversations. If a suspicious conversation is detected, the server automatically reports it to a fraud hotline, and the necessary measures are taken.

[1080] Announcement at the start of a call

[1081] When a user starts a call, the device plays an announcement saying, "This conversation is being recorded." This announcement acts as a deterrent to fraudulent activity on the other end of the call.

[1082] Speech to text

[1083] During a call, voice data is collected by the user's device and sent to a server, which then uses a voice recognition service to convert the voice data into text in real time, thereby recording the contents of the call as text information.

[1084] Text Analysis

[1085] The server inputs the real-time text of the call into a generative model for analysis. The generative model has an algorithm that detects specific keywords and phrases and identifies suspicious conversations. For example, if keywords such as "deposit," "PIN number," or "transfer" are detected, it will determine that there is a possibility of fraud.

[1086] Automatic reporting of suspicious conversations

[1087] If the generative model detects a suspicious conversation, the server automatically reports it to a fraud hotline. The report includes the content of the call and the generated warning message. The hotline then contacts the user, confirms the situation, and takes any necessary measures.

[1088] Managing the Exclusion List

[1089] Users can also exclude specific callers from detection. The device has an exclusion list management function that allows users to add the IDs of specific callers (for example, family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system, so the system can be operated without interfering with important calls.

[1090] Specific examples

[1091] Consider a case where an elderly user receives a fraudulent phone call, claiming to be an "urgent call from the bank," and is tricked into making a deposit. When the call begins, the device announces, "This conversation is being recorded," and audio data is collected. This audio data is converted into text in real time and analyzed by a generative model on a server. If the generative model detects keywords such as "deposit" and "bank" and determines that the conversation is suspicious, the server automatically reports the incident to a fraud hotline. The hotline contacts the user, confirms the situation, and takes necessary measures to prevent fraud before it occurs.

[1092] In this way, the present invention provides a system that can detect suspicious conversations in real time and respond quickly, thereby effectively preventing fraud against elderly people and others.

[1093] The processing flow will be explained below.

[1094] Step 1:

[1095] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," to deter the other party from engaging in fraudulent activity.

[1096] Step 2:

[1097] The device collects the audio data and sends the audio stream to the server. The audio stream is collected in real time to cover the entire call.

[1098] Step 3:

[1099] The server receives the audio stream and uses a speech recognition service to convert the audio data into text in real time, resulting in a written record of the conversation.

[1100] Step 4:

[1101] The server inputs the transcribed conversations into a generative model, which uses natural language processing (NLP) algorithms to analyze the text and detect specific keywords and phrases.

[1102] Step 5:

[1103] The generative model detects suspicious conversations. For example, if a conversation contains keywords such as "deposit," "PIN number," or "transfer," it determines that it may be fraudulent.

[1104] Step 6:

[1105] If the server detects a suspicious conversation, it automatically reports it to a fraud hotline, which includes the call details along with an automatically generated warning message.

[1106] Step 7:

[1107] The hotline receives the report and contacts the user. The hotline checks the situation and advises the user on the necessary measures (for example, freezing the account or contacting the police).

[1108] Step 8:

[1109] If the user wants to exclude certain callers from detection, the device provides an exclusion list management function. The user can add the IDs of specific callers (e.g., family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system.

[1110] Example 1

[1111] 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."

[1112] In recent years, the number of victims of telephone fraud has been increasing. Socially vulnerable people, such as the elderly, are often targeted, creating a need for an efficient system to prevent this from happening. However, conventional call monitoring systems lack the ability to detect suspicious conversations in real time, making it difficult to respond quickly. Furthermore, monitoring all calls poses the risk of privacy violations and interruption of important calls. Furthermore, systems lack the appropriate announcement function to enhance fraud deterrence.

[1113] 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.

[1114] In this invention, the server includes means for converting call content into text in real time, means for analyzing the converted call content using a generative model to detect suspicious conversations, means for automatically reporting to a reporting hotline when a suspicious conversation is detected, means for the terminal to play an announcement when a call starts, and means for managing a list of specific callers to be excluded from detection. This enables real-time monitoring and analysis of call content, and makes it possible to quickly report suspicious conversations while protecting privacy at the same time.

[1115] "Means of converting call content into text in real time" refers to technology that instantly converts voice data during a call into text information.

[1116] A "generative model" refers to an algorithm that uses machine learning to extract patterns from data and generate new information.

[1117] "Means for detecting suspicious conversations" refers to technology that analyzes the text content of phone calls and identifies conversations that may be fraudulent based on pre-set keywords and patterns.

[1118] "Means for automatically reporting to a reporting hotline" refers to a system that automatically sends the necessary information to a designated hotline when a suspicious conversation is detected.

[1119] "Means for the terminal to play an announcement when a call is initiated" refers to a function that automatically plays a pre-recorded message when a user initiates a call.

[1120] "Means for managing a list of specific callers to be excluded from detection" refers to a function for registering and managing callers that should not be monitored by the system.

[1121] The present invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model. Specifically, it detects suspicious conversations and automatically reports them to a hotline if necessary, thereby preventing fraud. The system configuration and operation are described in detail below.

[1122] System configuration

[1123] This system consists of the following main components:

[1124] 1. Terminal

[1125] The device through which a user makes a call, including smartphones and landlines.

[1126] The device is responsible for collecting voice data and sending it to the server.

[1127] It has the ability to play an announcement at the start of a call saying "This conversation is being recorded."

[1128] 2. Server

[1129] The received voice data is converted into text in real time.

[1130] The textual data is input into a generative AI model to detect suspicious conversations.

[1131] If a suspicious conversation is detected, a report will be automatically sent to the reporting hotline.

[1132] The generative AI model used is, for example, OpenAI's GPT-4.

[1133] 3. Reporting Desk

[1134] It has the function of receiving reports from the server, contacting the relevant user, and taking necessary measures.

[1135] Hardware and Software Examples

[1136] Speech recognition service: Converts voice data into text using services such as Google Cloud Speech-to-Text API and Amazon Transcribe.

[1137] Generative AI model: Analyzes the text of phone calls using models such as OpenAI's GPT-4.

[1138] Communication protocol: The HTTPS protocol is used to send audio data from the device to the server.

[1139] System operation explanation

[1140] 1. Start a call

[1141] When a user starts a call, the device automatically plays an announcement saying, "This conversation is being recorded."

[1142] 2. Collection and transmission of voice data

[1143] A microphone installed on the device collects voice data during the call and transmits it to the server in real time.

[1144] 3. Speech-to-text

[1145] The server uses a speech recognition service to convert the received voice data into text in real time.

[1146] 4. Text Data Analysis

[1147] The transcribed conversations are fed into a generative AI model that detects specific keywords and phrases to identify suspicious conversations.

[1148] For example, input the following prompt into the generative model:

[1149] Text of the call: "This is an urgent call from your bank. Please deposit money into XXX immediately."

[1150] Keywords to detect: deposit, bank, transfer, PIN, emergency

[1151] 5. Detecting and reporting suspicious conversations

[1152] When the generative model detects suspicious keywords such as "deposit" or "bank," the server automatically sends the relevant call details and a warning message to the reporting hotline.

[1153] 6. Responses via the reporting desk

[1154] The reporting hotline will contact the user in question, confirm the possibility of fraud, and take the necessary measures.

[1155] Specific examples

[1156] Consider a case where an elderly user receives a fraudulent phone call and is tricked into making a deposit, claiming to be an "urgent call from the bank." As the call begins, the device plays an announcement saying, "This conversation is being recorded," which is expected to have a deterrent effect on fraudulent activity. Collected voice data is converted into text in real time and analyzed by a generative model on the server. If the generative model detects suspicious keywords and determines that the conversation is suspicious, the server automatically notifies the hotline. The hotline then contacts the user and takes action to prevent fraud.

[1157] In this way, the present invention provides a system that effectively prevents fraud against elderly people and others by monitoring and analyzing call content in real time and immediately reporting any suspicious content.

[1158] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1159] Step 1: User initiates a call

[1160] A user initiates a call using a device. Specifically, the user launches a phone app, enters the other party's phone number, and presses the call button. This operation establishes the call.

[1161] Input: The phone number of the person you want to call

[1162] Output: Established call session

[1163] Step 2: The device plays an announcement

[1164] The device will play an announcement saying, "This conversation is being recorded." Specifically, the device will play an announcement message from its built-in speaker to let the other party know that the call is being recorded.

[1165] Input: Established call session

[1166] Output: Announcement message played

[1167] Step 3: The device collects audio data

[1168] The device collects voice data in real time during a call. Specifically, the device's microphone captures the voice of the call and stores the data in an internal buffer.

[1169] Input: Audio during a call

[1170] Output: Collected audio data

[1171] Step 4: The device sends the audio data to the server

[1172] The device sends the collected voice data to the server in real time. Specifically, the device uploads the voice data to the server using the HTTPS protocol.

[1173] Input: Collected audio data

[1174] Output: Audio data sent to the server

[1175] Step 5: The server converts the audio data into text

[1176] The server converts the received voice data into text in real time using a voice recognition service. Specifically, the server sends the voice data to a voice recognition API (e.g., Google Cloud Speech-to-Text API) and obtains the converted text.

[1177] Input: Received audio data

[1178] Output: Text of the call

[1179] Step 6: The server inputs the text data into the generative model

[1180] The server inputs the textual content of the call into a generative AI model, which analyzes it. Specifically, the server inputs the text data into a generative AI model (e.g., OpenAI's GPT-4) and generates prompts to detect suspicious keywords and phrases.

[1181] Input: Transcribed call

[1182] Output: Suspicious keywords and phrases detected

[1183] Step 7: The server detects suspicious conversations

[1184] The server detects suspicious conversations based on the output of the generative model. Specifically, the server analyzes the detection results from the generative model and determines whether specific keywords (e.g., "deposit," "bank," or "transfer") are included.

[1185] Input: Detection result

[1186] Output: Suspicious conversation detection flag

[1187] Step 8: The server automatically notifies the reporting point

[1188] If a suspicious conversation is detected, the server automatically reports it to the reporting desk. Specifically, the server generates the report content (textualized call content and a warning message) and sends a POST request to the reporting desk's API endpoint.

[1189] Input: Suspicious conversation detection flag, textual content of the call

[1190] Output: Report data to reporting desk

[1191] Step 9: The hotline contacts the user

[1192] The reporting desk will contact the relevant user, confirm the possibility of fraud, and take necessary measures. Specifically, the reporting desk operator will contact the user by phone or email to confirm the details of the situation.

[1193] Input: Report data

[1194] Output: User contact data

[1195] Step 10: Your device manages the exclusion list

[1196] To exclude specific callers from detection, the user manages an exclusion list on the device. Specifically, the user adds specific phone numbers to the exclusion list from the device's settings screen, and the device synchronizes the exclusion list with the server.

[1197] Input: Phone number of the person you want to add

[1198] Output: Updated exclusion list

[1199] (Application example 1)

[1200] 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."

[1201] While systems exist that monitor phone call content and detect suspicious conversations, they lack the ability to respond in real time or to quickly detect and report suspicious conversations in real-world environments. Furthermore, there is no mechanism for immediately notifying security personnel, making it difficult to prevent fraud and criminal activity. The present invention aims to solve these problems by providing a system that monitors and detects suspicious conversations in real time and quickly reports and notifies users.

[1202] 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.

[1203] In this invention, the server includes means for converting call content into text in real time, means for analyzing the converted call content using a generative model to detect suspicious conversations, means for automatically reporting to a fraud hotline when a suspicious conversation is detected, means for capturing the conversation content in real time using speech recognition and analyzing the suspicious conversation using a generative model, and means for sending an alert to a security officer when a suspicious conversation is detected. This makes it possible to monitor suspicious conversations in real time and respond quickly to prevent fraud and criminal acts.

[1204] "Means for converting call content into text in real time" refers to a function that converts voice data during a call into text format in real time, thereby recording the content of the call as text information.

[1205] "Means of analyzing textual content of phone calls using a generative model to detect suspicious conversations" refers to a function that uses a generative AI model to analyze textual content of phone calls and identify conversations that may involve fraudulent activity.

[1206] "Means of automatically reporting to a fraud hotline when a suspicious conversation is detected" refers to a function that automatically reports to a specialized agency responsible for eliminating fraud when a suspicious conversation is detected.

[1207] "Means of capturing conversation content in real time using voice recognition and analyzing suspicious conversations using a generative model" refers to a function that uses voice recognition technology to capture conversation content in real time, analyzes that data using a generative AI model, and identifies suspicious content.

[1208] "Means for sending alerts to security personnel when suspicious conversations are detected" refers to a function that quickly sends an alert to security personnel when suspicious conversations are discovered.

[1209] "A means of playing an announcement at the start of a call saying 'This conversation is being recorded'" refers to a function that plays a voice announcement at the start of a call to notify the other party that the conversation will be recorded.

[1210] "Means for managing a list that excludes conversations of specific callers from detection targets" refers to a function that excludes conversations of callers specified by the user from monitoring targets and manages that list.

[1211] A "reporting hotline" refers to a specialized agency or department that receives contact when suspicious conversations are detected and takes appropriate action.

[1212] A "generative model" refers to an AI algorithm or machine learning model that analyzes input data and generates new information or results, such as GPT-3 or BERT.

[1213] "Sending an alert" refers to the action of sending a warning or notification to a pre-defined notification destination when a problem occurs.

[1214] The present invention relates to a system that monitors telephone conversations in real time and quickly detects and reports suspicious conversations. The system configuration and processing flow for implementing the present invention will be described in detail below.

[1215] System Configuration

[1216] This system is broadly composed of the following three components:

[1217] 1. User terminal: refers to devices such as smart glasses, smartphones, head-mounted displays, etc. These devices have the ability to capture voice data in real time and transmit that data to a server.

[1218] 2. Server: This server has speech recognition capabilities to receive voice data and convert it to text, as well as the ability to analyze the text data using a generative AI model.

[1219] 3. Reporting system: This system sends an alert to a fraud reporting hotline or security officer if the server detects any suspicious conversations.

[1220] Program processing

[1221] User terminal

[1222] The user device captures the audio at the beginning of the call and plays an announcement saying, "This conversation is being recorded." The audio data is transmitted to the server in real time. The user device can perform this function using, for example, smart glasses or a smartphone.

[1223] server

[1224] The server receives the voice data sent from the user's device and converts it into text using speech recognition technology. Commonly used technologies on the server include Google Speech-to-Text API and IBM Watson.

[1225] The transcribed data is then fed into a generative AI model (e.g., GPT-3 or BERT), which has algorithms that identify specific keywords and patterns in the text to detect suspicious conversations.

[1226] Example of an input prompt for a generative AI model:

[1227] "Please analyze the following conversation. If there are any suspicious or dangerous statements, please mark them as suspicious. Conversation: 'We are going to give this confidential information to our business partner.'"

[1228] Reporting System

[1229] If a suspicious conversation is detected, the server automatically sends an alert to a fraud hotline or security officer, including the text of the suspicious conversation and any necessary warning messages.

[1230] Specific examples

[1231] For example, imagine a person in a confidential company meeting says, "I'm going to give this confidential information to a third party." This conversation is captured by smart glasses, and the audio data is sent to a server. The audio is converted into text in real time and analyzed by a generative AI model. If the model detects suspicious keywords such as "confidential information" or "give to a third party," an alert is immediately sent to security personnel. This rapid reporting prevents fraudulent activity before it occurs.

[1232] Through the system configuration and processing described above, the present invention realizes early detection of suspicious conversations in real time and rapid response.

[1233] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1234] Step 1:

[1235] At the start of a call, the user terminal captures the audio data and plays an announcement saying, "This conversation is being recorded."

[1236] Input: Call start trigger

[1237] Output: Announcement audio, captured audio data

[1238] How it works: Your smart glasses or smartphone will start recording audio through the microphone and play back an announcement.

[1239] Step 2:

[1240] The captured audio data is sent to the server in real time.

[1241] Input: Captured audio data

[1242] Output: Audio data sent to the server

[1243] Operation: The user terminal uses the communication module to send voice data to the server.

[1244] Step 3:

[1245] The server converts the received voice data into text using voice recognition technology.

[1246] Input: Audio data received on the server

[1247] Output: Text of the call

[1248] How it works: The server converts the audio data into text using the Google Speech-to-Text API or IBM Watson.

[1249] Step 4:

[1250] Using a generative AI model, the text of phone calls is analyzed to detect suspicious conversations.

[1251] Input: Transcribed call transcript

[1252] Output: Suspicious conversation detection result (suspicious / normal)

[1253] How it works: The server feeds text data into a generative AI model (e.g., GPT-3 or BERT) to analyze it for relevant keywords and patterns.

[1254] Step 5:

[1255] If a suspicious conversation is detected, the server sends an alert to a fraud hotline or security officer.

[1256] Input: Suspicious conversation detection results and corresponding text content

[1257] Output: Informational and warning messages

[1258] How it works: The server notifies security personnel via email or a dedicated application.

[1259] Step 6:

[1260] If the user wants to exclude a caller from detection, they can add and manage the ID of the specific caller to the exclusion list.

[1261] Input: User-specified callee ID

[1262] Output: Updated exclusion list

[1263] Operation: To exclude the user's specified caller from detection, the ID is sent to the server's exclusion list and the list is updated.

[1264] 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.

[1265] This invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model and an emotion engine. It detects suspicious conversations and automatically reports them to a fraud hotline as needed, preventing fraud before it happens. The system's program processing is explained in detail below.

[1266] System Overview

[1267] In this system, when a user initiates a call, the device collects voice data and the server converts the data into text in real time. In addition, the server analyzes the text of the call using a generative model and an emotion engine to detect suspicious conversations. If a suspicious conversation is detected, the server automatically reports it to a fraud hotline and the necessary measures are taken.

[1268] Announcement at the start of a call

[1269] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," which acts as a deterrent to fraudulent activity on the other end of the call.

[1270] Speech to text

[1271] During a call, voice data is collected by the user's device and sent to a server, which then uses a voice recognition service to convert the voice data into text in real time, thereby recording the contents of the call as text information.

[1272] Text Analysis

[1273] The server inputs the real-time text of the call into a generative model for analysis. The generative model uses natural language processing (NLP) algorithms to analyze the text and detect specific keywords and phrases. In addition, an emotion engine recognizes the user's emotions and incorporates that emotional information into the analysis results.

[1274] The role of the emotional engine

[1275] The emotion engine analyzes emotions from the user's tone of voice and facial expressions during a call and uses this information to determine suspicious conversations detected by the generative model. Specifically, if the emotion engine detects that the user is feeling strong anxiety or confusion, the generative model will take this emotional information into account and determine that there is a high possibility of fraud.

[1276] Automatic reporting of suspicious conversations

[1277] When the generative model and emotion engine work together to detect a suspicious conversation, the server automatically reports it to a fraud hotline. The report includes the content of the call, the generated warning message, and the user's emotional state. The hotline then contacts the user, confirms the situation, and takes necessary measures.

[1278] Managing the Exclusion List

[1279] Users can also exclude specific callers from detection. The device has an exclusion list management function that allows users to add the IDs of specific callers (for example, family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system, so the system can be operated without interfering with important calls.

[1280] Specific examples

[1281] For example, consider the case where an elderly user receives a fraudulent phone call, claiming to be an "urgent call from the bank" and requesting a deposit. When the user starts the call, the device announces, "This conversation is being recorded," and voice data is collected. This voice data is converted into text in real time and analyzed on the server using a generative model and emotion engine. If the generative model detects keywords such as "deposit" and "bank," and the emotion engine detects high anxiety as the user's emotional state, the server determines this to be a suspicious conversation and automatically reports it to a fraud hotline. The hotline contacts the user, confirms the situation, and instructs them to contact the bank or the police, for example, thereby preventing fraud before it occurs.

[1282] In this way, the present invention provides a system that can detect suspicious conversations in real time and respond quickly to them, effectively preventing fraud against elderly people and other victims. By combining it with an emotion engine, it becomes possible to detect suspicious conversations with even higher accuracy.

[1283] The processing flow will be explained below.

[1284] Step 1:

[1285] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," to deter the other party from engaging in fraudulent activity.

[1286] Step 2:

[1287] The device collects the audio data and sends the audio stream to the server. The audio stream is collected in real time to cover the entire call.

[1288] Step 3:

[1289] The server receives the audio stream and uses a speech recognition service to convert the audio data into text in real time, resulting in a written record of the conversation.

[1290] Step 4:

[1291] The server inputs the transcribed conversations into a generative model, which then uses natural language processing algorithms to analyze the text and detect specific keywords and phrases.

[1292] Step 5:

[1293] The server recognizes the user's emotions using an emotion engine, which analyzes the user's emotional state from the tone, rhythm, and speed of the voice.

[1294] Step 6:

[1295] The server combines the analysis results of the generative model and the emotion engine to determine whether the conversation is suspicious. For example, if keywords such as "deposit" and "PIN number" are detected and the emotion engine determines that the user is anxious or confused, it will determine that the conversation is suspicious.

[1296] Step 7:

[1297] If the generative model and emotion engine detect a suspicious conversation, the server automatically reports it to a fraud hotline, including the call content, a warning message, and the user's emotional state.

[1298] Step 8:

[1299] The hotline receives the report and contacts the user. The hotline checks the situation and advises the user on the necessary measures (for example, freezing the account or contacting the police).

[1300] Step 9:

[1301] If the user wants to exclude certain callers from detection, the device provides an exclusion list management function. The user can add the IDs of specific callers (e.g., family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system.

[1302] Step 10:

[1303] After the exclusion list management is applied, the device determines whether a new call is subject to detection based on the list. Calls with parties registered on the list are excluded from the analysis of the generative model and emotion engine and are treated as normal calls.

[1304] Example 2

[1305] 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."

[1306] In recent years, telephone fraud has been on the rise, with scams targeting the elderly in particular becoming more diverse. Under these circumstances, there is a need for a system that can monitor phone conversations in real time and quickly detect and report suspicious conversations. However, conventional systems face challenges in not only converting and analyzing phone conversations into text, but also in making highly accurate judgments that take into account the user's emotions. Another issue is the lack of functionality for flexibly managing the targets of monitoring so as not to interfere with routine or important calls.

[1307] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1308] In this invention, the server includes a means for converting call content into text in real time, a means for analyzing the converted call content using a generative model to detect suspicious conversations, a means for automatically reporting to a fraud hotline when a suspicious conversation is detected, and a means for analyzing the user's emotions using an emotion analysis engine and reflecting the results in determining whether the conversation is suspicious. This enables highly accurate detection and prompt reporting of suspicious conversations that take into account not only the call content but also the user's emotions. Furthermore, by including a means for announcing "This conversation is being recorded" at the start of a call and a means for managing a list of specific callers to be excluded from detection, it is possible to minimize interference with everyday calls while increasing the deterrent effect on the caller.

[1309] "Call content" refers to the voice information uttered by the user during a call.

[1310] "Real-time" is a time concept that refers to processing occurring almost simultaneously with the moment a call is made.

[1311] "Texting" is the process of converting audio information into written information.

[1312] A "generative model" is an algorithm that performs natural language processing and analyzes input text data to extract specific information.

[1313] "Suspicious conversations" refer to conversations that contain keywords or phrases that could be fraudulent.

[1314] An "emotion analysis engine" is software for analyzing a user's emotional state, detecting emotions from the tone and content of the voice during a call.

[1315] A "fraud reporting hotline" refers to an institution or service that accepts reports when suspicious conversations are detected.

[1316] An "announcement" is a voice message played from the terminal at the start of a call, informing the other party that the call will be recorded.

[1317] An "exclusion list" is a list in which the IDs of specific callers are registered so that calls with those people are not subject to monitoring.

[1318] The present invention is a system that converts phone conversations into text in real time and analyzes the converted text using a generative model and an emotion analysis engine. It detects suspicious conversations and automatically reports them to a fraud hotline as needed, thereby preventing fraud damage. A specific embodiment of the system is described below.

[1319] System Overview

[1320] In this system, when a user initiates a call, the device collects voice data, and the server converts the voice data into text in real time. In addition, the server analyzes the text of the call using a generative model and an emotion analysis engine to detect suspicious conversations. If a suspicious conversation is detected, the server automatically reports it to a fraud hotline, and necessary measures are taken.

[1321] Announcement at the start of a call

[1322] When a user starts a call, the device will play an announcement saying, "This conversation is being recorded," which acts as a deterrent to fraudulent activity on the other end of the call.

[1323] Speech to text

[1324] During a call, the user's device collects and transmits voice data to a server, which then uses a speech recognition service (e.g., Google Cloud Speech-to-Text) to convert the data into text in real time, thereby recording the contents of the call as text information.

[1325] Text Analysis

[1326] The server inputs the real-time text of the call into a generative model (e.g., OpenAI GPT-4) for analysis. The generative model uses natural language processing (NLP) algorithms to analyze the text and detect specific keywords and phrases. In addition, a sentiment analysis engine (e.g., IBM Watson Tone Analyzer) recognizes the user's emotions and incorporates that emotional information into the analysis results.

[1327] The role of sentiment analysis engines

[1328] The emotion analysis engine analyzes emotions from the tone and content of the user's voice during a call and uses this information to determine suspicious conversations detected by the generative model. Specifically, if the emotion analysis engine detects that the user is feeling strong anxiety or confusion, the generative model will take this emotional information into account and determine that there is a high possibility of fraud.

[1329] Automatic reporting of suspicious conversations

[1330] When the generative model and sentiment analysis engine work together to detect a suspicious conversation, the server automatically reports it to a fraud hotline. The report includes the content of the call, the generated warning message, and the user's emotional state. The hotline then contacts the user to confirm the situation and take necessary measures.

[1331] Managing the Exclusion List

[1332] Users can also exclude specific callers from detection. The device has an exclusion list management function that allows users to add the IDs of specific callers (for example, family members or acquaintances) to the exclusion list. Callers registered on this list will not be monitored by the system, so the system can be operated without interfering with important calls.

[1333] Specific examples

[1334] For example, consider the case where an elderly user receives a fraudulent phone call, claiming to be an "urgent call from the bank" and requesting a deposit. When the user starts the call, the device announces, "This conversation is being recorded," and voice data is collected. This voice data is converted into text in real time and analyzed on the server using a generative model and emotion analysis engine. If the generative model detects keywords such as "deposit" and "bank," and the emotion analysis engine detects high anxiety as the user's emotional state, the server determines this to be a suspicious conversation and automatically reports it to a fraud hotline. The hotline contacts the user, confirms the situation, and instructs them to contact the bank or the police, for example, thereby preventing the fraud from occurring.

[1335] Prompt Sentence Examples

[1336] "The user is on a call and the other party claims to be making an urgent call from their bank, requesting a deposit. The user's emotional state indicates high anxiety. Please determine whether this call is likely to be fraudulent."

[1337] The purpose of this system is to effectively prevent fraud, especially among the elderly, by detecting suspicious conversations in real time and responding quickly. By combining it with an emotion analysis engine, it is possible to detect suspicious conversations with even higher accuracy.

[1338] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1339] Step 1:

[1340] When a user starts a call, the device plays an announcement saying, "This conversation is being recorded." Specifically, the device's audio playback function is used to play a pre-recorded announcement. This announcement acts as a deterrent to the other party.

[1341] Step 2:

[1342] The terminal collects voice data uttered by the user during a call. This voice data is captured through the terminal's microphone and temporarily stored in buffer memory. The input is the user's voice, and the output is the voice data stored in the buffer memory.

[1343] Step 3:

[1344] The collected voice data is sent to the server in real time. Data is sent from the terminal to the server via network communication. The input is the voice data in the buffer memory, and the output is the voice data received on the server.

[1345] Step 4:

[1346] The server converts the received voice data into text using a voice recognition service. Specifically, it calls a voice recognition API such as Google Cloud Speech-to-Text and converts the voice data into text data. The input is the voice data, and the output is the text of the conversation.

[1347] Step 5:

[1348] The server inputs the converted conversation data into a generative model for analysis. The generative model uses natural language processing algorithms such as OpenAI GPT-4. The input is text data, and the output is the analysis results (e.g., detection of specific keywords or phrases).

[1349] Step 6:

[1350] At the same time, the server uses an emotion analysis engine to analyze the user's emotions. Specifically, it uses IBM Watson Tone Analyzer to extract emotions from the content of the call and the tone of the voice. The input is text data, and the output is emotional data.

[1351] Step 7:

[1352] The analysis results of the generative model are combined with the emotional data from the emotion analysis engine to identify suspicious conversations. The server then inputs this data into the algorithm to determine whether there is a possibility of fraud. The inputs are the analysis results and emotional data, and the output is the judgment result.

[1353] Step 8:

[1354] If a suspicious conversation is detected, the server automatically reports it to a fraud hotline. The report includes the text of the conversation, a generated warning message, and the user's emotional state. The input is the judgment result and related data, and the output is a report message.

[1355] Step 9:

[1356] Users manage an exclusion list to exclude specific callers from detection. Users can add the IDs of specific callers to the exclusion list through their devices. This information is sent from the device to the server. The input is the user's operation, and the output is the updated exclusion list.

[1357] (Application example 2)

[1358] 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."

[1359] In recent years, the number of fraudulent phone calls has been increasing, with elderly people and corporate employees being particularly targeted. When receiving a suspicious call, it is currently difficult to quickly and accurately assess the content and take appropriate measures. Furthermore, conventional methods for analyzing call content have low accuracy and there is a risk of suspicious conversations being overlooked, so an effective system to prevent fraudulent phone calls is needed.

[1360] 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.

[1361] In this invention, the server includes means for converting the contents of the call into text in real time, means for analyzing the converted contents of the call using a generative model to detect suspicious conversations, means for analyzing the emotional state during the call in real time using an emotion analysis engine, and means for automatically reporting to a fraud hotline if a suspicious conversation is detected. This makes it possible to analyze the contents of the call in real time, detect suspicious conversations with high accuracy, and take prompt measures.

[1362] "Means of converting phone conversations into text in real time" refers to a system that uses voice recognition technology to instantly convert conversations that take place during a phone call into text information.

[1363] "Means for detecting suspicious conversations through analysis using a generative model" is a system that analyzes the text content of conversations using a generative model based on a natural language processing algorithm to identify suspicious conversations that may involve fraud or dishonesty.

[1364] An "emotion analysis engine" refers to a program or device that evaluates emotions from the tone of voice and linguistic expressions during a call and analyzes that information in real time.

[1365] "Means for automatically reporting to fraud hotlines" refers to a system that automatically notifies fraud prevention officers and agencies when suspicious conversations are detected.

[1366] "Means for playing announcements" refers to a system that has the function of automatically playing a notification voice message such as "This conversation is being recorded" at the start of a call.

[1367] "Means for managing a list that excludes conversations of specific callers from detection" refers to a database or program that registers and manages information about pre-designated callers in order to exclude their conversations from monitoring.

[1368] The present invention is a system that converts telephone conversations into text in real time and analyzes the converted text using a generative model and an emotion analysis engine. The system detects suspicious conversations and automatically reports them to a fraud hotline, thereby preventing fraud damage. The specific configuration and operation of the system for implementing the present invention are described below.

[1369] System configuration

[1370] 1. Device:

[1371] The device has the ability to record the contents of calls and plays an announcement at the start of the call saying, "This conversation is being recorded." The recorded voice data is converted into text in real time and sent to a server.

[1372] 2. Server:

[1373] The server has the following main functions:

[1374] Speech recognition engine: Using the Python speech_recognition library, recorded voice data is converted into text in real time. This function allows the contents of calls to be instantly recorded as text information on the server.

[1375] Generative Model: Using Natural Language Processing (NLP) algorithms, Hugging Face's transformers library is used to analyze the transcribed conversations. This generative model detects specific keywords and phrases associated with fraud and scams.

[1376] Emotion analysis engine: Evaluates emotions from speech tone and linguistic expressions. This allows the user's emotional state to be analyzed in real time and reflected in the generative model's decisions.

[1377] Automatic reporting function: If a suspicious conversation is detected, a notification containing the call content and analysis results will be automatically sent to a fraud reporting hotline. This function uses the Python requests library.

[1378] Specific examples

[1379] For example, consider the case where an elderly user receives a fraudulent phone call requesting that the user deposit money into their bank account. When the user starts the call, the device plays a message saying, "This conversation is being recorded," and begins collecting voice data. This voice data is converted into text in real time and sent to a server.

[1380] The server uses a speech recognition engine to convert voice data into text and inputs the text content into a generative model for analysis. It also uses an emotion analysis engine to evaluate the user's emotional state, and if high anxiety is detected, it reflects this in the generative model's judgment. Specifically, it detects keywords such as "deposit" and "bank" from the text content, and if it determines that the user is in a very anxious state, the server recognizes this as a suspicious conversation.

[1381] The automated reporting function then automatically sends a notification containing the call details and analysis results to the fraud hotline, allowing the hotline to respond quickly and prevent damage before it occurs.

[1382] Prompt Sentence Examples

[1383] "Enter the text of the call below to analyze it for potential fraud."

[1384] Call: "Please deposit money into my bank account."

[1385] Emotional state: "Very anxious"

[1386] In this way, the system of the present invention provides an effective means of preventing fraud by detecting suspicious conversations in real time with high accuracy and responding quickly.By combining a speech recognition engine, a generative model, and a sentiment analysis engine, we can build a very powerful security service.

[1387] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1388] Step 1:

[1389] When a user starts a call, the device plays an announcement saying "This conversation is being recorded." This serves as a warning to the other party and a deterrent to fraudulent activity. The input is the call start event, and the output is the playback of the announcement audio.

[1390] Step 2:

[1391] The terminal starts collecting voice data in real time during the call. The voice data is obtained from the microphone and stored in a buffer. The input is the continuous voice data from the microphone, and the output is the collected voice data buffer.

[1392] Step 3:

[1393] The collected voice data is sent from the device to the server. The input is the data buffer of the device, and the output is the voice data transferred to the server. Specifically, the data is sent using the device's communication module.

[1394] Step 4:

[1395] The server converts the received voice data into text in real time using a speech recognition engine. A speech recognition engine (e.g., Python's speech_recognition library) is used to generate text data from the input voice data. The input is voice data, and the output is the text of the conversation.

[1396] Step 5:

[1397] The server inputs the converted conversation into a generative model (e.g., Hugging Face's transformers library) and analyzes it using natural language processing algorithms to detect specific keywords and phrases related to fraud. The input is text data, and the output is the results of keyword detection.

[1398] Step 6:

[1399] The server uses an emotion analysis engine to evaluate the text and emotional state in real time. The emotion analysis engine evaluates emotions from speech tone and linguistic expressions and incorporates this information into the analysis results of the generative model. The input is audio data and text data, and the output is the detected emotional state.

[1400] Step 7:

[1401] The server integrates the results of the generative model and the sentiment analysis engine to determine whether suspicious conversations have been detected. The inputs are the keyword detection results from the generative model and the evaluation results from the sentiment analysis engine, and the output is the detection results of suspicious conversations.

[1402] Step 8:

[1403] If a suspicious conversation is detected, the server automatically sends a notification to the fraud hotline, including the call content and analysis results. The notification also includes the user's emotional state. The input is the detection result of the suspicious conversation, and the output is a notification message to the fraud hotline. Specifically, an HTTP request is sent using the Python requests library.

[1404] Step 9:

[1405] The server manages a list of users to exclude conversations with specific callers from detection. It stores the information of the callers specified by the user in a database and manages detection targets based on that list. The input is the list of callers specified by the user, and the output is the updated exclusion list.

[1406] This series of processes enables the system to quickly and accurately detect suspicious conversations and prevent fraud before it occurs.

[1407] 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.

[1408] 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.

[1409] 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 robot 414.

[1410] 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.

[1411] 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.

[1412] 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.

[1413] 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).

[1414] 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.

[1415] 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."

[1416] 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.

[1417] 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).

[1418] 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.

[1419] 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.

[1420] 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.

[1421] 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.

[1422] 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.

[1423] 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.

[1424] 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.

[1425] 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.

[1426] 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.

[1427] 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.

[1428] The following is further disclosed regarding the above embodiment.

[1429] (Claim 1)

[1430] A means of converting call content into text in real time,

[1431] A means of analyzing the text of phone calls using a generative model to detect suspicious conversations;

[1432] A means to automatically report suspicious conversations to a fraud hotline if they are detected, and

[1433] A system including:

[1434] (Claim 2)

[1435] In fraud detection call systems,

[1436] Further including a means to play an announcement at the beginning of a call stating "This conversation is being recorded";

[1437] 10. The system of claim 1.

[1438] (Claim 3)

[1439] The system further includes a means for managing a list for excluding conversations of specific call parties from detection targets.

[1440] 10. The system of claim 1.

[1441] The above are examples of claims containing the key technical features that form the heart and definition of the invention.

[1442] "Example 1"

[1443] (Claim 1)

[1444] A means of converting call content into text in real time,

[1445] A means of analyzing the text of phone calls using a generative model to detect suspicious conversations;

[1446] A means to automatically report suspicious conversations to a reporting hotline when they are detected;

[1447] a means for the terminal to play an announcement at the start of a call;

[1448] A means for managing a list of specific callers to be excluded from detection;

[1449] A system including:

[1450] (Claim 2)

[1451] 10. The system of claim 1, further comprising: means for playing an announcement at the beginning of a call that says "This conversation is being recorded."

[1452] (Claim 3)

[1453] 10. The system of claim 1, further comprising means for managing a list of specific parties to be excluded from detection.

[1454] "Application Example 1"

[1455] (Claim 1)

[1456] A means of converting call content into text in real time,

[1457] A means of analyzing the text of phone calls using a generative model to detect suspicious conversations;

[1458] A means to automatically report suspicious conversations to a fraud hotline if they are detected, and

[1459] A means of capturing conversation content in real time using speech recognition and analyzing suspicious conversations using a generative model;

[1460] A means of sending alerts to security personnel if suspicious conversations are detected;

[1461] A system including:

[1462] (Claim 2)

[1463] Further including a means to play an announcement at the beginning of a call stating "This conversation is being recorded";

[1464] 10. The system of claim 1.

[1465] (Claim 3)

[1466] The system further includes a means for managing a list for excluding conversations of specific call parties from detection targets.

[1467] 10. The system of claim 1.

[1468] "Example 2: Combining Emotion Engines"

[1469] (Claim 1)

[1470] A means of converting call content into text in real time,

[1471] A means of analyzing the text of phone calls using a generative model to detect suspicious conversations;

[1472] A means to automatically report suspicious conversations to a fraud hotline if they are detected, and

[1473] A means for analyzing a user's emotions using an emotion analysis engine and reflecting the results in determining whether a conversation is suspicious;

[1474] A system including:

[1475] (Claim 2)

[1476] Further including a means to play an announcement at the beginning of a call stating "This conversation is being recorded";

[1477] 10. The system of claim 1.

[1478] (Claim 3)

[1479] The system further includes a means for managing a list for excluding conversations of specific call parties from detection targets.

[1480] 10. The system of claim 1.

[1481] "Application example 2 when combining emotion engines"

[1482] (Claim 1)

[1483] A means of converting call content into text in real time,

[1484] A means of analyzing the text of phone calls using a generative model to detect suspicious conversations;

[1485] means for analyzing the emotional state during a call in real time using an emotion analysis engine;

[1486] A means to automatically report suspicious conversations to a fraud hotline if they are detected, and

[1487] A system including:

[1488] (Claim 2)

[1489] Further including a means to play an announcement at the beginning of a call stating "This conversation is being recorded";

[1490] 10. The system of claim 1.

[1491] (Claim 3)

[1492] The system further includes a means for managing a list for excluding conversations of specific call parties from detection targets.

[1493] 10. The system of claim 1. [Explanation of symbols]

[1494] 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 of converting call content into text in real time, A means of analyzing the text of phone calls using a generative model to detect suspicious conversations; A means to automatically report suspicious conversations to a fraud hotline if they are detected, and A system including:

2. In fraud detection call systems, Further including a means to play an announcement at the beginning of a call stating "This conversation is being recorded"; The system of claim 1 .

3. The system further includes a means for managing a list for excluding conversations of specific call parties from detection targets. The system of claim 1 .

Images