system

Abstract

Provide a system. A means for recording audio; a means for converting the recorded audio into text; a means for summarizing the converted characters; a means for analyzing the summarized text and assessing the security of the communication; means for controlling call connection based on the evaluation result; A system including:

Description

[Technical Field]

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

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

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

[0004] Telephone scams and other fraudulent activities often target the elderly, causing serious damage. Senior citizens are vulnerable to fraudulent techniques and are at risk of losing their personal information and assets. Therefore, effective measures to prevent such fraud and protect the elderly are needed. [Means for solving the problem]

[0005] The present invention provides a system that records phone conversations, converts the audio data into text, then summarizes it and evaluates the safety of the call using generative AI. This system first includes a means for recording audio and a means for converting the recorded audio into text. Next, it includes a means for summarizing the converted text and a means for analyzing the summarized text to evaluate the safety of the call. Finally, it includes a means for controlling call connections based on the evaluation results, thereby protecting elderly people from fraud. Specifically, the system is configured to include a means for reporting to a fraud prevention department if the evaluation results indicate that the call is unsafe.

[0006] "Audio" refers to sounds, including human voices, obtained through a telephone, microphone, etc.

[0007] "Recording" refers to the act of saving data such as calls and voice messages and making it reusable at a later date.

[0008] "Convert to text" refers to the process of converting audio data into text data.

[0009] "Summarization" refers to the act of concisely summarizing the original text or audio content and extracting only the important information.

[0010] "Analysis" refers to the process of examining and evaluating the converted text data in detail to understand its content and characteristics.

[0011] "Communication Safety Assessment" refers to an assessment of a call to determine whether the content of the call is related to fraud or fraudulent activity.

[0012] "Evaluation results" refer to conclusions or findings regarding safety obtained based on the analysis.

[0013] "Controlling call connections" refers to the act of allowing or denying a call connection based on the evaluation results.

[0014] A "means" refers to a method, device, or process used to achieve a particular end.

[0015] "Fraud response department" refers to an organization or department that specializes in taking measures against and responding to fraudulent activities. [Brief explanation of the drawings]

[0016] [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

[0017] 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.

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

[0019] 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).

[0020] 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.

[0021] 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.

[0022] 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.

[0023] 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."

[0024] [First embodiment]

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

[0026] 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.

[0027] 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).

[0028] 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.

[0029] 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.

[0030] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form 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.

[0031] 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.

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

[0033] 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.

[0034] 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.

[0035] 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.

[0036] 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."

[0037] The present invention provides a system that records the contents of phone calls, converts the voice data into text, then summarizes it, and evaluates the safety of the call using a generation AI. Below, we will explain the program processing and specific examples of this system.

[0038] System configuration and processing

[0039] 1. Call logging

[0040] When the user receives a call, the server uses a call recording means to capture the user's voice through a microphone, thereby obtaining the contents of the call as audio data.

[0041] 2. Transcription of audio data

[0042] The server uses the acquired voice data to activate a means for converting voice to text, which is performed using voice recognition technology to obtain the contents of the call as text data.

[0043] 3. Summary of Contents

[0044] The server then uses a means for summarizing the converted text data to concisely summarize the contents of the text, allowing the user to obtain the necessary information in a condensed form.

[0045] 4. AI-powered filtering

[0046] The server analyzes the summarized text and uses generative AI to assess the safety of the call, identifying whether the call is related to fraud or other fraudulent activity.

[0047] 5. Call connection control

[0048] The server activates a means to control the call connection based on the AI ​​evaluation results. Only if the evaluation result is judged to be "safe" will the call be connected to the user's device.

[0049] If the evaluation result is judged to be "risky," the server will not connect the call to the user's device and will report it to the fraud department.

[0050] Specific examples

[0051] For example, consider the case where an elderly person receives a phone call from someone claiming to be from their bank. The call is first recorded by a server. The recorded voice data is then converted into text. The converted text might say something like, "We've recently observed unusual transactions from your account. Please provide your personal information to verify."

[0052] The server summarizes this text into a short summary such as "Abnormal account transaction confirmation, request for personal information." The summarized text is then evaluated by a generative AI, which in this case determines it to be "risky."

[0053] As a result, the server will not connect the call to the user's device but will instead report it to the fraud department. This system significantly reduces the risk of seniors becoming victims of telephone fraud.

[0054] The processing flow will be explained below.

[0055] Step 1:

[0056] The user answers the call. The server activates the call recording function and captures the user's voice in real time through the microphone. This voice data is stored as is on the server.

[0057] Step 2:

[0058] The server analyzes the recorded voice data and activates a voice-to-text converter. Specifically, it uses voice recognition technology to convert the voice data into text data. The result of this conversion is a text representation of the contents of the call.

[0059] Step 3:

[0060] The server uses a means to summarize the converted text data, turning long conversations into concise summaries. The summarization process extracts important information and summarizes it in an easy-to-understand format.

[0061] Step 4:

[0062] The server then invokes the generative AI to analyze the summarized text. The AI ​​uses natural language processing techniques to analyze the text and assess whether the call is related to fraud or other misconduct, using historical data and pattern recognition.

[0063] Step 5:

[0064] The server receives the evaluation result from the AI ​​generator and activates a means to control the call connection based on the result. If the evaluation result is judged to be "safe," the server connects the call to the user's device.

[0065] Step 6:

[0066] If the evaluation results in a "risk" rating, the server will not connect the call to the user's device and will instead report it to the fraud department. This report will be sent along with the specific content of the call, allowing for a prompt response.

[0067] Example 1

[0068] 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."

[0069] In the past, elderly people and users with low security awareness have been at high risk of falling victim to telephone fraud and fraudulent calls. To solve this problem and ensure the safety of calls, a system is needed that can carefully examine the content of calls and determine their safety in real time.

[0070] 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.

[0071] In this invention, the server includes a means for recording calls, a means for converting the recorded call voice into text, a means for summarizing the converted text, a means for analyzing the summarized text using a generative AI model and evaluating the safety of the call content, and a means for controlling the call connection based on the evaluation result. This makes it possible to reduce the risk of elderly people and users with low security awareness becoming victims of telephone fraud and fraudulent calls.

[0072] "Call" refers to the act of communicating for the purpose of voice communication.

[0073] "Recording" refers to the act or means of saving audio data.

[0074] "Convert to text" refers to the process of converting voice data into text data.

[0075] "Summarization" refers to the process of shortening long text data, extracting important information, and summarizing it concisely.

[0076] A "generative AI model" refers to an algorithm or program that uses artificial intelligence to generate and analyze data.

[0077] "Analysis" refers to the process of examining data and understanding its properties and patterns.

[0078] "Evaluating call content for safety" refers to the evaluation process to determine whether the content of a call is related to fraud or fraudulent activity.

[0079] "Controlling call connections" refers to the process of allowing or denying call connections based on specific evaluation results.

[0080] "Surveillance and Response Unit" refers to a unit dedicated to responding to fraud and other misconduct when it occurs.

[0081] This invention provides a system that records the contents of phone calls, converts the voice data into text, then summarizes it and evaluates the safety of the call using generative AI. When a user answers a call, the server processes the call contents in real time, providing a safe calling environment.

[0082] 1. Call logging

[0083] When a user answers a call, the server automatically starts recording the call. Specifically, the server captures audio data in real time through the device's microphone and saves it as audio data. The saved audio file format is typically WAV or MP3.

[0084] 2. Transcription of audio data

[0085] The server uses speech recognition software such as Google® Cloud Speech-to-Text API to convert the recorded voice data into text data. The server sends the audio file to the API, which processes it and obtains the text data. The obtained text data is a direct transcription of the contents of the call.

[0086] 3. Summary of Contents

[0087] The server uses natural language processing techniques such as the Transformers library to summarize the retrieved text data. The text data is input into a summarization model to obtain a summary sentence. For example, the text data "Hello, this is a message from the bank" is converted into a summary sentence such as "A message from the bank."

[0088] 4. AI-powered filtering

[0089] The server uses a generative AI model (e.g., GPT-4 (registered trademark)) to evaluate the safety of the call content based on the summary text. A prompt text is generated for the generative AI model, and the summary text is input into the prompt text to obtain the evaluation result. An example of a prompt text is, "Please evaluate the content of the following call. Summary text: 'Confirmation of abnormal account transactions, request for personal information'. Is this call safe or dangerous?"

[0090] 5. Call connection control

[0091] The server manages the call connection based on the evaluation results of the generated AI model. The call is connected to the user's device only if the evaluation result is judged to be "safe." If the result is judged to be "dangerous," the call connection is cut off and a report is sent to the monitoring department. This process ensures the safety of the user.

[0092] Specific examples

[0093] For example, if an elderly person receives a call from someone claiming to be from their bank, the call is first recorded by the server. The recorded voice data is converted into text data using the Google Cloud Speech-to-Text API. The converted text may then say something like, "We have recently observed unusual transactions in your account. Please provide your personal information to confirm." This text data is then summarized using the Transformers library as "Confirming unusual account transactions, requesting personal information."

[0094] The summarized text is fed into the GPT-4 model, and the generative AI model determines that the call is "risky." As a result, the server does not connect the call to the user's device and reports it to the surveillance department. This significantly reduces the risk of seniors becoming victims of telephone fraud.

[0095] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0096] Step 1:

[0097] When a user answers a call, the server starts recording the call. Specifically, the server captures audio data in real time from the device's microphone and saves it as an audio file, typically in WAV or MP3 format.

[0098] Input: User's call audio

[0099] Output: Audio file (WAV or MP3 format)

[0100] Step 2:

[0101] The server uses speech recognition software (e.g., Google Cloud Speech-to-Text API) to convert the recorded audio file into text. The server sends the audio file to the API, which processes it and obtains the text data.

[0102] Input: Audio file

[0103] Output: Text data

[0104] Step 3:

[0105] The server uses natural language processing techniques (e.g., the Transformers library) to summarize the acquired text data. The summarization model narrows down the input text data to the important information, shortens it, and generates a summary sentence.

[0106] Input: Text data

[0107] Output: Summary

[0108] Step 4:

[0109] The server evaluates the safety of the call content using a generative AI model (e.g., GPT-4) based on the summary. Specifically, the server generates a prompt for the generative AI model, inputs the summary along with this prompt, and obtains the evaluation result. An example of a prompt is, "Please evaluate the following call content. Summary: 'Confirmation of abnormal account transactions, request for personal information'. Is this call safe or dangerous?"

[0110] Input: Summary, prompt

[0111] Output: Evaluation result (safe / dangerous)

[0112] Step 5:

[0113] The server controls the call connection based on the evaluation results obtained by the generative AI model. The call is connected to the user's device only if the evaluation result is judged to be "safe." If the evaluation result is judged to be "dangerous," the call connection is cut off and a notification is sent to the monitoring and response department. Specifically, the server controls the communication software according to the evaluation result and issues a warning to the security department if necessary.

[0114] Input: Evaluation result

[0115] Output: Connect or disconnect the call, notify the monitoring department

[0116] (Application example 1)

[0117] 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."

[0118] In recent years, telephone fraud and malicious communications targeting the elderly have been increasing, creating a need to protect users from such fraudulent activities. However, current call monitoring systems have difficulty detecting fraud in real time, and there is also concern that legitimate calls may be blocked due to false positives. Therefore, it is necessary to develop a system that can accurately and quickly evaluate the security of calls and detect fraudulent calls in real time.

[0119] 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.

[0120] In this invention, the server includes a means for recording voice, a means for converting voice to text, a means for summarizing the text, a means for analyzing the summarized text using a generative AI model and evaluating the safety of communication, a means for controlling call connections based on the evaluation result, and a means for reporting to an anti-fraud department if the evaluation is unsafe. This makes it possible to evaluate the safety of calls quickly and with high accuracy, and to detect and deal with fraudulent calls in real time.

[0121] "Audio recording means" is a mechanism for collecting audio data during a call and storing it in a digital format.

[0122] "Means for converting voice to text" refers to technology that analyzes recorded voice data and converts its contents into text data.

[0123] The "text summarization means" is an algorithm that concisely summarizes the converted text data and extracts only the necessary information.

[0124] A "generative AI model" is a trained model that uses artificial intelligence to analyze data and make inferences.

[0125] "Means for assessing the security of communications" refers to the process of using generative AI models to determine whether a call contains fraud or scam content.

[0126] The "means for controlling call connection" is a function that determines whether to connect a call to a user terminal or block the call based on the evaluation result.

[0127] "Means to report to fraud prevention department" is a function that notifies a specialized department if the evaluated content of a call is determined to be fraudulent.

[0128] The present invention relates to a system for monitoring telephone call content in real time to detect fraud and other illegal activities. Specific embodiments of this system will be described below.

[0129] First, when a user answers a call, the server uses the device's microphone to record the call audio. This audio data is then saved in a digital format. Next, speech recognition technology is used to convert the audio data into text. Specifically, this technology uses Google's speech recognition API.

[0130] The converted text data is then further summarized. For summarization, we use the Hugging Face transformers library. As a concrete example, we use the T5 model to concisely summarize the text content.

[0131] The next step is to use a generative AI model to assess the safety of the summarized text. This assessment is performed using a custom SafetyEvaluator model. The generative AI model is fed a specific prompt, such as "Please assess whether this call is fraud-related. Provide the following summarized text: 'Confirm unusual account activity, request for personal information.'"

[0132] After the evaluation result is obtained, the call connection is controlled based on the result. If the evaluation result is determined to be "safe," the call is successfully connected to the user's device. Conversely, if the evaluation result is "unsafe," the call is blocked and automatically reported to the fraud department. The report contains the necessary information to enable the specialized department to take action quickly.

[0133] In this way, the present invention makes it possible to accurately and quickly evaluate the security of calls and detect and deal with fraudulent calls in real time, thereby protecting users from fraudulent activities and significantly reducing the damage caused by telephone fraud and other crimes.

[0134] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0135] Step 1:

[0136] When a user receives a call, the device's microphone is used to record the call audio. The call audio data is transmitted from the microphone to a server in real time and stored in a digital format. The input is raw audio data, and the output is digital audio data.

[0137] Step 2:

[0138] The server converts the recorded voice data into text data using speech recognition technology. This process uses Google's speech recognition API. The input is digital voice data and the output is text data. The server analyzes the voice data and converts each voice segment into the corresponding text.

[0139] Step 3:

[0140] The server summarizes the converted text data. This summarization utilizes Hugging Face's transformers library, specifically the T5 model. The input is the converted text data, and the output is the summary text. The server analyzes the text data, extracts important information, and summarizes it concisely.

[0141] Step 4:

[0142] The server analyzes the summarized text data and evaluates safety using a generative AI model. Using a custom SafetyEvaluator model, the input is the summarized text and the output is the safety evaluation result. For example, this process inputs a prompt such as, "Please evaluate whether this call is fraud-related. Provide the following summarized text: 'Confirm abnormal account transactions, request for personal information.'"

[0143] Step 5:

[0144] The server controls the call connection based on the evaluation results. If the evaluation result is judged to be "safe," the call is connected to the user's device. The input is the safety evaluation result, and the output is whether the call can be connected. Conversely, if the call is judged to be "unsafe," the call is blocked and an automatic report is sent to the fraud prevention department. The report includes a summary of the specific call content and details of the evaluation result.

[0145] These processing steps enable accurate and rapid assessment of call security and real-time detection and response to fraudulent calls.

[0146] 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.

[0147] The present invention provides a system that records the contents of phone calls, converts the voice data into text, then summarizes it, and evaluates the safety of the call using a generative AI and emotion engine. Below, we will explain the program processing and specific examples of this system.

[0148] System configuration and processing

[0149] 1. Call logging

[0150] When the user answers the phone, the server uses a call recording method to capture the user's voice through a microphone, and this voice data is stored on the server as is.

[0151] 2. Transcription of audio data

[0152] The server uses the acquired voice data to activate a means for converting voice to text, which is performed using voice recognition technology to obtain the contents of the call as text data.

[0153] 3. Summary of Contents

[0154] The server then uses a means for summarizing the converted text data to concisely summarize the contents of the text, allowing the user to obtain the necessary information in a condensed form.

[0155] 4. AI-powered filtering

[0156] The server then invokes the generative AI to analyze the summarized text. The AI ​​uses natural language processing techniques to analyze the text and assess whether the call is related to fraud or other misconduct, using historical data and pattern recognition.

[0157] 5. Emotion Recognition by Emotion Engine

[0158] The server uses the user's voice data to activate the emotion engine, which uses voice analysis technology to assess the user's emotions and stress level, thereby detecting emotions such as confusion or nervousness.

[0159] 6. Integration of evaluation results and call connection control

[0160] The server combines the evaluation results of the generative AI and the emotion engine to perform a comprehensive safety assessment of the call content. If the assessment result is deemed "safe," the server connects the call to the user's device.

[0161] If the evaluation result is judged to be "risky," the server will not connect the call to the user's device and will report it to the fraud department.

[0162] Specific examples

[0163] For example, consider the case where an elderly person receives a phone call from someone claiming to be from a bank. The call is first recorded by a server. The recorded voice data is then converted into text. The converted text might say something like, "We've recently observed unusual transactions from your account. Please provide your personal information to verify."

[0164] The server summarizes this text into a short summary such as "Abnormal account transaction confirmation, request for personal information." The summarized text is then evaluated by a generative AI, which in this case determines it to be "risky."

[0165] At the same time, the emotion engine analyzes the voice data and determines whether the user is confused or nervous. This information is also added to the overall evaluation. As a result, the server will not connect the call to the user's device and will report it to the fraud department.

[0166] This system can significantly reduce the risk of elderly people becoming victims of telephone fraud. In addition, by taking into account the user's emotions, it can provide a more accurate evaluation.

[0167] The processing flow will be explained below.

[0168] Step 1:

[0169] The user answers the call. The server activates the call recording function and captures the user's voice in real time through the microphone. This voice data is stored as is on the server.

[0170] Step 2:

[0171] The server uses the recorded voice data to activate a voice-to-text conversion means. Specifically, it uses voice recognition technology to convert the voice data into text data. The result of this conversion is a text representation of the contents of the call.

[0172] Step 3:

[0173] The server uses a means to summarize the converted text data, turning long conversations into concise summaries. The summarization process extracts important information and summarizes it in an easy-to-understand format.

[0174] Step 4:

[0175] The server then invokes the generative AI to analyze the summarized text. The AI ​​uses natural language processing techniques to analyze the text and assess whether the call is related to fraud or other misconduct, using historical data and pattern recognition.

[0176] Step 5:

[0177] The server receives the evaluation results from the AI ​​generator, along with the user's voice data, which then activates the emotion engine. The emotion engine uses voice analysis technology to evaluate the user's emotions and stress level, detecting whether the user is confused, nervous, or relieved.

[0178] Step 6:

[0179] The server combines the results of the generative AI and the emotion engine to perform a comprehensive safety assessment of the call content, specifically determining whether the likelihood of fraud is high or low, and whether the user is nervous.

[0180] Step 7:

[0181] If the evaluation result is deemed "safe," the server connects the call to the user's device, allowing the user to continue the call with peace of mind.

[0182] Step 8:

[0183] If the evaluation results in a "risk" rating, the server will not connect the call to the user's device and will instead report it to the fraud department. This report will include the specific content of the call and information about the user's emotional state, allowing for a prompt response.

[0184] Specific examples

[0185] For example, consider the case where an elderly person receives a phone call from someone claiming to be from their bank. The call is first recorded by a server. The recorded voice data is then converted into text. The converted text might say something like, "We've recently observed unusual transactions from your account. Please provide your personal information to verify."

[0186] The server summarizes this text into a short summary such as "Abnormal account transaction confirmation, request for personal information." The summarized text is then evaluated by a generative AI, which in this case determines it to be "risky."

[0187] At the same time, the emotion engine analyzes the voice data and evaluates whether the user is confused or nervous. This information is also taken into account in the overall evaluation. As a result, the server reports the call to the fraud department without connecting it to the user's device. This system can significantly reduce the risk of elderly people becoming victims of telephone fraud. Furthermore, taking the user's emotions into account allows for more accurate evaluation.

[0188] Example 2

[0189] 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."

[0190] In recent years, telephone fraud targeting the elderly has been increasing, and an effective system to prevent such fraud is required. To prevent such fraud, it is necessary to analyze the contents of calls quickly and accurately and to identify fraudulent or dangerous calls in real time. Conventional systems do not take into account the user's emotions or stress state, making it difficult to improve accuracy.

[0191] 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.

[0192] In this invention, the server includes means for recording voice, means for converting the recorded voice into text, means for summarizing the converted text, means for analyzing the summarized text using a generative AI model to assess the risk of fraud or misconduct, means for analyzing the recorded voice using an emotion analysis engine to assess the emotional state, means for integrating the evaluation results of the generative AI model and the emotion analysis engine to comprehensively assess the safety of communication, and means for controlling call connections based on the evaluation results. This makes it possible to assess the risk of telephone fraud with high accuracy and comprehensively determine the safety of calls while taking the user's emotional state into consideration.

[0193] "Audio recording means" refers to devices or techniques used to capture the contents of telephone conversations.

[0194] The "means for converting recorded voice into text" refers to a voice recognition technology for converting voice data into text data.

[0195] The "means for summarizing converted characters" is a technique for concisely summarizing text data that has been recognized by speech recognition.

[0196] "Means for analyzing using a generative AI model to assess the risk of fraud or misconduct" refers to a technology that uses AI technology to analyze text data and assess the possibility of fraud or misconduct.

[0197] The "means for analyzing using an emotion analysis engine and assessing the emotional state" is a technology that analyzes voice data to assess the user's emotions and stress state.

[0198] "Means for integrating evaluation results and comprehensively assessing the safety of communication" refers to a technology that integrates the evaluation results obtained from a generative AI model and an emotion analysis engine to comprehensively judge the safety of a call.

[0199] The "means for controlling call connection based on the evaluation result" is a technology for controlling call connection based on the safety evaluation result of the call content.

[0200] This invention relates to a system that records phone conversations, converts the audio data into text, then summarizes it and evaluates the safety of the call using generative AI and an emotion engine.

[0201] First, when a user answers a phone call, the contents of the call are recorded by the device. The recorded voice data is immediately sent to the server and stored on the server side. Specifically, the user's voice is picked up through the device's microphone, and the voice data is transferred to the server using a network protocol (e.g., UDP or TCP). The server then stores this data in cloud storage (e.g., Amazon S3).

[0202] Next, the server converts the recorded voice data into text data using voice recognition technology, specifically a voice recognition engine (e.g., Google Cloud Speech-to-Text API). The server sends the voice file to this API and stores the returned text data in a database (e.g., MySQL (registered trademark)).

[0203] The server then uses a summarization engine (e.g., OpenAI® GPT-3®) to summarize the converted text data. The server sends the text data along with the prompt to the summarization engine and receives the summarized text data. The summary results are also stored in a database.

[0204] The server then uses a generative AI model to perform a risk assessment on the summarized text. Specifically, the summarized text is sent to a generative AI (e.g., OpenAI GPT-3) again to assess the likelihood of fraud or misconduct. The results of this assessment are also stored on the server.

[0205] Furthermore, the server launches an emotion analysis engine (e.g., IBM Watson (registered trademark) Tone Analyzer) to recognize and evaluate the user's emotions from the recorded voice data. The server sends the voice data to the emotion analysis engine and receives the analysis results. These evaluation results are also stored in the database.

[0206] Finally, the server combines the results of the generative AI model and the sentiment analysis engine to evaluate the overall safety of the call. If the evaluation result is "safe," the server connects the call to the user's device. On the other hand, if the evaluation result is "risky," the server notifies the fraud department and blocks the call.

[0207] Specific examples

[0208] For example, if an elderly person receives a call from someone claiming to be from a bank, the system operates as follows:

[0209] 1. When a user receives a call, the device captures the voice data and sends it to the server.

[0210] 2. The server sends the audio data to the Google Cloud Speech-to-Text API and converts it into text data.

[0211] 3. The converted text data is summarized using OpenAI GPT-3. For example, a message such as "We have recently observed unusual transactions from your account. Please provide your personal information to confirm" is summarized as "Confirming unusual transactions on your account, requesting the provision of personal information."

[0212] 4. The server analyzes this summary text using OpenAI GPT-3 and rates it as "dangerous."

[0213] 5. At the same time, IBM Watson Tone Analyzer is used to analyze the user's emotions and assess whether the user is confused or nervous.

[0214] 6. The server combines these evaluation results and, if it determines the call is "risky," it reports it to the fraud department and blocks the call.

[0215] Prompt Sentence Examples

[0216] "If you receive a call from someone claiming to be a bank employee and they ask for personal information, transcribe the call and then summarize it to assess its security."

[0217] The present invention allows for safe and reliable call management by effectively assessing the risk of telephone fraud and taking into account the emotional state of the user.

[0218] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0219] Step 1:

[0220] When a user answers a call and the device starts a call, the server captures and receives the voice data in real time. Specifically, the device's microphone captures the user's voice and sends the voice data to the server using UDP or TCP protocols. The server then stores the received voice data in storage (e.g., Amazon S3).

[0221] Input: User's voice data

[0222] Output: Audio file saved to storage

[0223] Step 2:

[0224] The server retrieves the voice data from the storage and converts it into text data using voice recognition technology. It then activates a voice recognition engine (e.g., Google Cloud Speech-to-Text API) and sends the voice data to a cloud service. The converted text data is then stored in a database (e.g., MySQL).

[0225] Input: Audio file

[0226] Output: Character data stored in the database

[0227] Step 3:

[0228] The server retrieves text data from the database and condenses it using summarization technology. It then launches a summarization engine (e.g., OpenAI GPT-3) and sends the text data along with a prompt to the summarization engine. The summarized text data is then stored in the database.

[0229] Input: Character data

[0230] Output: Summary text stored in the database

[0231] Step 4:

[0232] The server analyzes the summarized text using a generative AI model to assess the risk of fraud and misconduct. The summarized text, along with a prompt, is sent to the generative AI (e.g., OpenAI GPT-3) to obtain a risk assessment result, which is also stored in a database.

[0233] Input: Summary text

[0234] Output: Risk assessment results stored in a database

[0235] Step 5:

[0236] The server passes the voice data to an emotion analysis engine to analyze the user's emotion. It then starts the emotion analysis engine (e.g., IBM Watson Tone Analyzer) and sends the voice data. The analysis results are stored in a database.

[0237] Input: Audio file

[0238] Output: Sentiment analysis results stored in a database

[0239] Step 6:

[0240] The server combines the evaluation results of the generative AI model and the sentiment analysis engine to perform a comprehensive safety assessment of the call content. The evaluation results are input into the integrated algorithm to perform a final safety assessment. If the evaluation result is judged to be "safe," the server instructs the call to be connected to the user's device. If the evaluation result is judged to be "risky," the server reports the call to the fraud department and blocks the call.

[0241] Input: Risk assessment results, sentiment analysis results

[0242] Output: Connect or block the call and report it to the fraud department

[0243] (Application example 2)

[0244] 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."

[0245] In modern society, telephone fraud is on the rise. The elderly and general public are especially at risk of falling prey to fraudulent calls, and traditional manual countermeasures have their limitations. There is also a need for systems that can automatically analyze the content of phone calls, evaluate their safety, and provide feedback and alerts to users.

[0246] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for recording voice, means for converting the recorded voice into text, means for summarizing the converted text, means for analyzing the summarized text using a generation AI and an emotion engine to evaluate the safety of communication and the user's emotions, means for controlling call connection based on the evaluation results, and means for feeding back the evaluation results to the user. This makes it possible to evaluate the safety of calls in real time, provide quick feedback to the user, and reduce the risk of fraud.

[0247] An "audio recording means" is a device or system that records telephone conversations in real time and stores them as digital audio data for subsequent processing.

[0248] A "means for converting recorded voice into text" is a device or system that uses voice recognition technology to convert recorded voice data into text data.

[0249] The "means for summarizing converted characters" is a device or system that analyzes text data, extracts important information, and summarizes it concisely.

[0250] "Means for analyzing using generative AI and an emotion engine to assess the safety of communication and the user's emotions" refers to a device or system that uses a generative AI model and an emotion analysis engine to assess emotions from the summarized text content and voice, and determine the safety of the call and the user's emotional state.

[0251] "Means for controlling call connections based on evaluation results" refers to a device or system that connects a call if it is judged to be safe based on the evaluation results of the generative AI and emotion engine, and blocks the connection if it is judged to be dangerous.

[0252] "Means for providing feedback of evaluation results to the user" refers to a device or system that notifies the user of the evaluation results by the generative AI and emotion engine, and provides information on the safety of the call and the connection status.

[0253] A "means for reporting to the fraud department" is a device or system that, if a call is deemed dangerous, notifies the fraud department of that information and prompts them to take appropriate action.

[0254] To implement this invention, the following system configuration and programs are required: The entire system is built around three elements: a server, a terminal, and a user.

[0255] System Configuration

[0256] 1. Server

[0257] As a means of recording audio, the server captures audio data during a call in real time and stores it as digital data.

[0258] To convert the recorded voice into text, the server uses voice recognition technology to convert the acquired voice data into text data. Specifically, it uses the speech_recognition library.

[0259] To summarize the converted text, the server uses natural language processing techniques to extract and summarize important information from the text data, for example, by using spacy to analyze the content of the text.

[0260] The summarized text is analyzed by the generative AI and emotion engine, and the generative AI model and emotion engine are activated as a means of evaluating the safety of communication and the user's emotions. The generative AI model uses a model from the transformers library, and the emotion engine uses the emotion_recognition library.

[0261] As a means of controlling call connections based on the evaluation results, the server summarizes the evaluation results from the AI ​​and emotion engine, and connects the call if it is judged to be safe, and stops the connection if it is judged to be dangerous.

[0262] As a means of providing feedback on the evaluation results to the user, the evaluation results are sent to the terminal and notified to the user.

[0263] 2. Terminal

[0264] The device receives the evaluation results sent from the server and provides visual or auditory feedback to the user, for example by displaying a warning message using the smartphone's notification function.

[0265] 3. Users

[0266] During a telephone call, the user receives feedback from the terminal and decides whether to continue or terminate the call as needed.

[0267] Specific examples

[0268] For example, if a user receives a call from someone claiming to be from a "bank," the call is first recorded by the server. The recorded voice data is converted into text data and recognized as, "We have recently observed unusual activity on your account. Please provide your personal information for verification." The server summarizes this text and summarizes it as, "Abnormal activity on account, request for personal information." This summarized text is then analyzed using a generative AI model and an emotion engine, and is rated as "dangerous." This evaluation result is sent to the device and displayed as a warning to the user. As a result, the user recognizes that the call may be fraudulent and can take appropriate action.

[0269] Prompt Sentence Examples

[0270] The following prompts are fed into the generative AI model for evaluation:

[0271] "Please summarize the following call and assess its likelihood of fraud or misconduct.\nCall content: {Call content}"

[0272] Thus, the present invention provides an effective system for reducing the risk of telephone fraud and keeping users safe.

[0273] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0274] Step 1:

[0275] When the server detects that the user has started a phone call, it starts recording the audio in real time by capturing the audio data using a microphone and storing the digital data on the server.

[0276] Input: User's call audio

[0277] Output: Digital audio data

[0278] Step 2:

[0279] The server converts the recorded voice data into text data using speech recognition technology. Specifically, it uses the speech_recognition library to convert voice to text.

[0280] Input: Digital audio data

[0281] Output: Text data

[0282] Step 3:

[0283] The server summarizes the text data using natural language processing techniques, such as the spacy library, to extract and summarize key information concisely.

[0284] Input: Text data

[0285] Output: Summarized text data

[0286] Step 4:

[0287] The server runs a generative AI model and an emotion engine to analyze the content of the summarized text and assess the safety of the call and the user's emotions. The generative AI model uses the transformers library, and the emotion engine uses the emotion_recognition library.

[0288] Input: Summarized text data

[0289] Output: Safety evaluation results and emotion evaluation results

[0290] Step 5:

[0291] The server controls the call connection based on the evaluation results of the generative AI and emotion engine. If the call is judged to be safe, the connection is maintained, and if it is judged to be dangerous, the connection is cut off.

[0292] Input: Safety evaluation results and emotion evaluation results

[0293] Output: Call connection status (connected / disconnected)

[0294] Step 6:

[0295] The server sends the evaluation results to the device and notifies the user, so the user can immediately know whether the call is safe or dangerous. The device displays this using its notification function.

[0296] Input: Safety evaluation results and emotion evaluation results

[0297] Output: User notification

[0298] Step 7:

[0299] The user receives feedback from the device and decides whether to continue or terminate the call, and if necessary, terminates the call and reports the call to the fraud department.

[0300] Input: User notification

[0301] Output: User behavior (continue / cancel call, report)

[0302] 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.

[0303] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0304] 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.

[0305] [Second embodiment]

[0306] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0307] 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.

[0308] 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).

[0309] 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.

[0310] 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.

[0311] 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).

[0312] 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.

[0313] 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.

[0314] 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.

[0315] 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.

[0316] 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.

[0317] 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."

[0318] The present invention provides a system that records the contents of phone calls, converts the voice data into text, then summarizes it, and evaluates the safety of the call using a generation AI. Below, we will explain the program processing and specific examples of this system.

[0319] System configuration and processing

[0320] 1. Call logging

[0321] When the user receives a call, the server uses a call recording means to capture the user's voice through a microphone, thereby obtaining the contents of the call as audio data.

[0322] 2. Transcription of audio data

[0323] The server uses the acquired voice data to activate a means for converting voice to text, which is performed using voice recognition technology to obtain the contents of the call as text data.

[0324] 3. Summary of Contents

[0325] The server then uses a means for summarizing the converted text data to concisely summarize the contents of the text, allowing the user to obtain the necessary information in a condensed form.

[0326] 4. AI-powered filtering

[0327] The server analyzes the summarized text and uses generative AI to assess the safety of the call, identifying whether the call is related to fraud or other fraudulent activity.

[0328] 5. Call connection control

[0329] The server activates a means to control the call connection based on the AI ​​evaluation results. Only if the evaluation result is judged to be "safe" will the call be connected to the user's device.

[0330] If the evaluation result is judged to be "risky," the server will not connect the call to the user's device and will report it to the fraud department.

[0331] Specific examples

[0332] For example, consider the case where an elderly person receives a phone call from someone claiming to be from their bank. The call is first recorded by a server. The recorded voice data is then converted into text. The converted text might say something like, "We've recently observed unusual transactions from your account. Please provide your personal information to verify."

[0333] The server summarizes this text into a short summary such as "Abnormal account transaction confirmation, request for personal information." The summarized text is then evaluated by a generative AI, which in this case determines it to be "risky."

[0334] As a result, the server will not connect the call to the user's device but will instead report it to the fraud department. This system significantly reduces the risk of seniors becoming victims of telephone fraud.

[0335] The processing flow will be explained below.

[0336] Step 1:

[0337] The user answers the call. The server activates the call recording function and captures the user's voice in real time through the microphone. This voice data is stored as is on the server.

[0338] Step 2:

[0339] The server analyzes the recorded voice data and activates a voice-to-text converter. Specifically, it uses voice recognition technology to convert the voice data into text data. The result of this conversion is a text representation of the contents of the call.

[0340] Step 3:

[0341] The server uses a means to summarize the converted text data, turning long conversations into concise summaries. The summarization process extracts important information and summarizes it in an easy-to-understand format.

[0342] Step 4:

[0343] The server then invokes the generative AI to analyze the summarized text. The AI ​​uses natural language processing techniques to analyze the text and assess whether the call is related to fraud or other misconduct, using historical data and pattern recognition.

[0344] Step 5:

[0345] The server receives the evaluation result from the AI ​​generator and activates a means to control the call connection based on the result. If the evaluation result is judged to be "safe," the server connects the call to the user's device.

[0346] Step 6:

[0347] If the evaluation results in a "risk" rating, the server will not connect the call to the user's device and will instead report it to the fraud department. This report will be sent along with the specific content of the call, allowing for a prompt response.

[0348] Example 1

[0349] 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."

[0350] In the past, elderly people and users with low security awareness have been at high risk of falling victim to telephone fraud and fraudulent calls. To solve this problem and ensure the safety of calls, a system is needed that can carefully examine the content of calls and determine their safety in real time.

[0351] 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.

[0352] In this invention, the server includes a means for recording calls, a means for converting the recorded call voice into text, a means for summarizing the converted text, a means for analyzing the summarized text using a generative AI model and evaluating the safety of the call content, and a means for controlling the call connection based on the evaluation result. This makes it possible to reduce the risk of elderly people and users with low security awareness becoming victims of telephone fraud and fraudulent calls.

[0353] "Call" refers to the act of communicating for the purpose of voice communication.

[0354] "Recording" refers to the act or means of saving audio data.

[0355] "Convert to text" refers to the process of converting voice data into text data.

[0356] "Summarization" refers to the process of shortening long text data, extracting important information, and summarizing it concisely.

[0357] A "generative AI model" refers to an algorithm or program that uses artificial intelligence to generate and analyze data.

[0358] "Analysis" refers to the process of examining data and understanding its properties and patterns.

[0359] "Evaluating call content for safety" refers to the evaluation process to determine whether the content of a call is related to fraud or fraudulent activity.

[0360] "Controlling call connections" refers to the process of allowing or denying call connections based on specific evaluation results.

[0361] "Surveillance and Response Unit" refers to a unit dedicated to responding to fraud and other misconduct when it occurs.

[0362] This invention provides a system that records the contents of phone calls, converts the voice data into text, then summarizes it and evaluates the safety of the call using generative AI. When a user answers a call, the server processes the call contents in real time, providing a safe calling environment.

[0363] 1. Call logging

[0364] When a user answers a call, the server automatically starts recording the call. Specifically, the server captures audio data in real time through the device's microphone and saves it as audio data. The saved audio file format is typically WAV or MP3.

[0365] 2. Transcription of audio data

[0366] The server uses speech recognition software such as Google Cloud Speech-to-Text API to convert the recorded voice data into text data. The server sends the audio file to the API, which processes it and obtains the text data. The obtained text data is a direct transcription of the contents of the call.

[0367] 3. Summary of Contents

[0368] The server uses natural language processing techniques such as the Transformers library to summarize the retrieved text data. The text data is input into a summarization model to obtain a summary sentence. For example, the text data "Hello, this is a message from the bank" is converted into a summary sentence such as "A message from the bank."

[0369] 4. AI-powered filtering

[0370] The server uses a generative AI model (e.g., GPT-4) to evaluate the safety of the call content based on the summary. It generates a prompt for the generative AI model and obtains the evaluation result by inputting the summary into the prompt. An example of a prompt is, "Please evaluate the following call content. Summary: 'Confirmation of abnormal account transactions, request for personal information'. Is this call safe or dangerous?"

[0371] 5. Call connection control

[0372] The server manages the call connection based on the evaluation results of the generated AI model. The call is connected to the user's device only if the evaluation result is judged to be "safe." If the result is judged to be "dangerous," the call connection is cut off and a report is sent to the monitoring department. This process ensures the safety of the user.

[0373] Specific examples

[0374] For example, if an elderly person receives a call from someone claiming to be from their bank, the call is first recorded by the server. The recorded voice data is converted into text data using the Google Cloud Speech-to-Text API. The converted text may then say something like, "We have recently observed unusual transactions in your account. Please provide your personal information to confirm." This text data is then summarized using the Transformers library as "Confirming unusual account transactions, requesting personal information."

[0375] The summarized text is fed into the GPT-4 model, and the generative AI model determines that the call is "risky." As a result, the server does not connect the call to the user's device and reports it to the surveillance department. This significantly reduces the risk of seniors becoming victims of telephone fraud.

[0376] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0377] Step 1:

[0378] When a user answers a call, the server starts recording the call. Specifically, the server captures audio data in real time from the device's microphone and saves it as an audio file, typically in WAV or MP3 format.

[0379] Input: User's call audio

[0380] Output: Audio file (WAV or MP3 format)

[0381] Step 2:

[0382] The server uses speech recognition software (e.g., Google Cloud Speech-to-Text API) to convert the recorded audio file into text. The server sends the audio file to the API, which processes it and obtains the text data.

[0383] Input: Audio file

[0384] Output: Text data

[0385] Step 3:

[0386] The server uses natural language processing techniques (e.g., the Transformers library) to summarize the acquired text data. The summarization model narrows down the input text data to the important information, shortens it, and generates a summary sentence.

[0387] Input: Text data

[0388] Output: Summary

[0389] Step 4:

[0390] The server evaluates the safety of the call content using a generative AI model (e.g., GPT-4) based on the summary. Specifically, the server generates a prompt for the generative AI model, inputs the summary along with this prompt, and obtains the evaluation result. An example of a prompt is, "Please evaluate the following call content. Summary: 'Confirmation of abnormal account transactions, request for personal information'. Is this call safe or dangerous?"

[0391] Input: Summary, prompt

[0392] Output: Evaluation result (safe / dangerous)

[0393] Step 5:

[0394] The server controls the call connection based on the evaluation results obtained by the generative AI model. The call is connected to the user's device only if the evaluation result is judged to be "safe." If the evaluation result is judged to be "dangerous," the call connection is cut off and a notification is sent to the monitoring and response department. Specifically, the server controls the communication software according to the evaluation result and issues a warning to the security department if necessary.

[0395] Input: Evaluation result

[0396] Output: Connect or disconnect the call, notify the monitoring department

[0397] (Application example 1)

[0398] 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."

[0399] In recent years, telephone fraud and malicious communications targeting the elderly have been increasing, creating a need to protect users from such fraudulent activities. However, current call monitoring systems have difficulty detecting fraud in real time, and there is also concern that legitimate calls may be blocked due to false positives. Therefore, it is necessary to develop a system that can accurately and quickly evaluate the security of calls and detect fraudulent calls in real time.

[0400] 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.

[0401] In this invention, the server includes a means for recording voice, a means for converting voice to text, a means for summarizing the text, a means for analyzing the summarized text using a generative AI model and evaluating the safety of communication, a means for controlling call connections based on the evaluation result, and a means for reporting to an anti-fraud department if the evaluation is unsafe. This makes it possible to evaluate the safety of calls quickly and with high accuracy, and to detect and deal with fraudulent calls in real time.

[0402] "Audio recording means" is a mechanism for collecting audio data during a call and storing it in a digital format.

[0403] "Means for converting voice to text" refers to technology that analyzes recorded voice data and converts its contents into text data.

[0404] The "text summarization means" is an algorithm that concisely summarizes the converted text data and extracts only the necessary information.

[0405] A "generative AI model" is a trained model that uses artificial intelligence to analyze data and make inferences.

[0406] "Means for assessing the security of communications" refers to the process of using generative AI models to determine whether a call contains fraud or scam content.

[0407] The "means for controlling call connection" is a function that determines whether to connect a call to a user terminal or block the call based on the evaluation result.

[0408] "Means to report to fraud prevention department" is a function that notifies a specialized department if the evaluated content of a call is determined to be fraudulent.

[0409] The present invention relates to a system for monitoring telephone call content in real time to detect fraud and other illegal activities. Specific embodiments of this system will be described below.

[0410] First, when a user answers a call, the server uses the device's microphone to record the call audio. This audio data is then saved in a digital format. Next, speech recognition technology is used to convert the audio data into text. Specifically, this technology uses Google's speech recognition API.

[0411] The converted text data is then further summarized. For summarization, we use the Hugging Face transformers library. As a concrete example, we use the T5 model to concisely summarize the text content.

[0412] The next step is to use a generative AI model to assess the safety of the summarized text. This assessment is performed using a custom SafetyEvaluator model. The generative AI model is fed a specific prompt, such as "Please assess whether this call is fraud-related. Provide the following summarized text: 'Confirm unusual account activity, request for personal information.'"

[0413] After the evaluation result is obtained, the call connection is controlled based on the result. If the evaluation result is determined to be "safe," the call is successfully connected to the user's device. Conversely, if the evaluation result is "unsafe," the call is blocked and automatically reported to the fraud department. The report contains the necessary information to enable the specialized department to take action quickly.

[0414] In this way, the present invention makes it possible to accurately and quickly evaluate the security of calls and detect and deal with fraudulent calls in real time, thereby protecting users from fraudulent activities and significantly reducing the damage caused by telephone fraud and other crimes.

[0415] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0416] Step 1:

[0417] When a user receives a call, the device's microphone is used to record the call audio. The call audio data is transmitted from the microphone to a server in real time and stored in a digital format. The input is raw audio data, and the output is digital audio data.

[0418] Step 2:

[0419] The server converts the recorded voice data into text data using speech recognition technology. This process uses Google's speech recognition API. The input is digital voice data and the output is text data. The server analyzes the voice data and converts each voice segment into the corresponding text.

[0420] Step 3:

[0421] The server summarizes the converted text data. This summarization utilizes Hugging Face's transformers library, specifically the T5 model. The input is the converted text data, and the output is the summary text. The server analyzes the text data, extracts important information, and summarizes it concisely.

[0422] Step 4:

[0423] The server analyzes the summarized text data and evaluates safety using a generative AI model. Using a custom SafetyEvaluator model, the input is the summarized text and the output is the safety evaluation result. For example, this process inputs a prompt such as, "Please evaluate whether this call is fraud-related. Provide the following summarized text: 'Confirm abnormal account transactions, request for personal information.'"

[0424] Step 5:

[0425] The server controls the call connection based on the evaluation results. If the evaluation result is judged to be "safe," the call is connected to the user's device. The input is the safety evaluation result, and the output is whether the call can be connected. Conversely, if the call is judged to be "unsafe," the call is blocked and an automatic report is sent to the fraud prevention department. The report includes a summary of the specific call content and details of the evaluation result.

[0426] These processing steps enable accurate and rapid assessment of call security and real-time detection and response to fraudulent calls.

[0427] 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.

[0428] The present invention provides a system that records the contents of phone calls, converts the voice data into text, then summarizes it, and evaluates the safety of the call using a generative AI and emotion engine. Below, we will explain the program processing and specific examples of this system.

[0429] System configuration and processing

[0430] 1. Call logging

[0431] When the user answers the phone, the server uses a call recording method to capture the user's voice through a microphone, and this voice data is stored on the server as is.

[0432] 2. Transcription of audio data

[0433] The server uses the acquired voice data to activate a means for converting voice to text, which is performed using voice recognition technology to obtain the contents of the call as text data.

[0434] 3. Summary of Contents

[0435] The server then uses a means for summarizing the converted text data to concisely summarize the contents of the text, allowing the user to obtain the necessary information in a condensed form.

[0436] 4. AI-powered filtering

[0437] The server then invokes the generative AI to analyze the summarized text. The AI ​​uses natural language processing techniques to analyze the text and assess whether the call is related to fraud or other misconduct, using historical data and pattern recognition.

[0438] 5. Emotion Recognition by Emotion Engine

[0439] The server uses the user's voice data to activate the emotion engine, which uses voice analysis technology to assess the user's emotions and stress level, thereby detecting emotions such as confusion or nervousness.

[0440] 6. Integration of evaluation results and call connection control

[0441] The server combines the evaluation results of the generative AI and the emotion engine to perform a comprehensive safety assessment of the call content. If the assessment result is deemed "safe," the server connects the call to the user's device.

[0442] If the evaluation result is judged to be "risky," the server will not connect the call to the user's device and will report it to the fraud department.

[0443] Specific examples

[0444] For example, consider the case where an elderly person receives a phone call from someone claiming to be from a bank. The call is first recorded by a server. The recorded voice data is then converted into text. The converted text might say something like, "We've recently observed unusual transactions from your account. Please provide your personal information to verify."

[0445] The server summarizes this text into a short summary such as "Abnormal account transaction confirmation, request for personal information." The summarized text is then evaluated by a generative AI, which in this case determines it to be "risky."

[0446] At the same time, the emotion engine analyzes the voice data and determines whether the user is confused or nervous. This information is also added to the overall evaluation. As a result, the server will not connect the call to the user's device and will report it to the fraud department.

[0447] This system can significantly reduce the risk of elderly people becoming victims of telephone fraud. In addition, by taking into account the user's emotions, it can provide a more accurate evaluation.

[0448] The processing flow will be explained below.

[0449] Step 1:

[0450] The user answers the call. The server activates the call recording function and captures the user's voice in real time through the microphone. This voice data is stored as is on the server.

[0451] Step 2:

[0452] The server uses the recorded voice data to activate a voice-to-text conversion means. Specifically, it uses voice recognition technology to convert the voice data into text data. The result of this conversion is a text representation of the contents of the call.

[0453] Step 3:

[0454] The server uses a means to summarize the converted text data, turning long conversations into concise summaries. The summarization process extracts important information and summarizes it in an easy-to-understand format.

[0455] Step 4:

[0456] The server then invokes the generative AI to analyze the summarized text. The AI ​​uses natural language processing techniques to analyze the text and assess whether the call is related to fraud or other misconduct, using historical data and pattern recognition.

[0457] Step 5:

[0458] The server receives the evaluation results from the AI ​​generator, along with the user's voice data, which then activates the emotion engine. The emotion engine uses voice analysis technology to evaluate the user's emotions and stress level, detecting whether the user is confused, nervous, or relieved.

[0459] Step 6:

[0460] The server combines the results of the generative AI and the emotion engine to perform a comprehensive safety assessment of the call content, specifically determining whether the likelihood of fraud is high or low, and whether the user is nervous.

[0461] Step 7:

[0462] If the evaluation result is deemed "safe," the server connects the call to the user's device, allowing the user to continue the call with peace of mind.

[0463] Step 8:

[0464] If the evaluation results in a "risk" rating, the server will not connect the call to the user's device and will instead report it to the fraud department. This report will include the specific content of the call and information about the user's emotional state, allowing for a prompt response.

[0465] Specific examples

[0466] For example, consider the case where an elderly person receives a phone call from someone claiming to be from their bank. The call is first recorded by a server. The recorded voice data is then converted into text. The converted text might say something like, "We've recently observed unusual transactions from your account. Please provide your personal information to verify."

[0467] The server summarizes this text into a short summary such as "Abnormal account transaction confirmation, request for personal information." The summarized text is then evaluated by a generative AI, which in this case determines it to be "risky."

[0468] At the same time, the emotion engine analyzes the voice data and evaluates whether the user is confused or nervous. This information is also taken into account in the overall evaluation. As a result, the server reports the call to the fraud department without connecting it to the user's device. This system can significantly reduce the risk of elderly people becoming victims of telephone fraud. Furthermore, taking the user's emotions into account allows for more accurate evaluation.

[0469] Example 2

[0470] 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."

[0471] In recent years, telephone fraud targeting the elderly has been increasing, and an effective system to prevent such fraud is required. To prevent such fraud, it is necessary to analyze the contents of calls quickly and accurately and to identify fraudulent or dangerous calls in real time. Conventional systems do not take into account the user's emotions or stress state, making it difficult to improve accuracy.

[0472] 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.

[0473] In this invention, the server includes means for recording voice, means for converting the recorded voice into text, means for summarizing the converted text, means for analyzing the summarized text using a generative AI model to assess the risk of fraud or misconduct, means for analyzing the recorded voice using an emotion analysis engine to assess the emotional state, means for integrating the evaluation results of the generative AI model and the emotion analysis engine to comprehensively assess the safety of communication, and means for controlling call connections based on the evaluation results. This makes it possible to assess the risk of telephone fraud with high accuracy and comprehensively determine the safety of calls while taking the user's emotional state into consideration.

[0474] "Audio recording means" refers to devices or techniques used to capture the contents of telephone conversations.

[0475] The "means for converting recorded voice into text" refers to a voice recognition technology for converting voice data into text data.

[0476] The "means for summarizing converted characters" is a technique for concisely summarizing text data that has been recognized by speech recognition.

[0477] "Means for analyzing using a generative AI model to assess the risk of fraud or misconduct" refers to a technology that uses AI technology to analyze text data and assess the possibility of fraud or misconduct.

[0478] The "means for analyzing using an emotion analysis engine and assessing the emotional state" is a technology that analyzes voice data to assess the user's emotions and stress state.

[0479] "Means for integrating evaluation results and comprehensively assessing the safety of communication" refers to a technology that integrates the evaluation results obtained from a generative AI model and an emotion analysis engine to comprehensively judge the safety of a call.

[0480] The "means for controlling call connection based on the evaluation result" is a technology for controlling call connection based on the safety evaluation result of the call content.

[0481] This invention relates to a system that records phone conversations, converts the audio data into text, then summarizes it and evaluates the safety of the call using generative AI and an emotion engine.

[0482] First, when a user answers a phone call, the contents of the call are recorded by the device. The recorded voice data is immediately sent to the server and stored on the server side. Specifically, the user's voice is picked up through the device's microphone, and the voice data is transferred to the server using a network protocol (e.g., UDP or TCP). The server then stores this data in cloud storage (e.g., Amazon S3).

[0483] The server then converts the recorded audio data into text data using speech recognition technology, specifically a speech recognition engine (e.g., Google Cloud Speech-to-Text API). The server sends the audio file to the API and stores the returned text data in a database (e.g., MySQL).

[0484] The server then uses a summarization engine (e.g., OpenAI GPT-3) to summarize the converted text data. The server sends the text data along with the prompt to the summarization engine and receives the summarized text data. The summary results are also stored in a database.

[0485] The server then uses a generative AI model to perform a risk assessment on the summarized text. Specifically, the summarized text is sent to a generative AI (e.g., OpenAI GPT-3) again to assess the likelihood of fraud or misconduct. The results of this assessment are also stored on the server.

[0486] Furthermore, the server launches an emotion analysis engine (e.g., IBM Watson Tone Analyzer) to recognize and evaluate the user's emotions from the recorded voice data. The server sends the voice data to the emotion analysis engine and receives the analysis results. These evaluation results are also stored in the database.

[0487] Finally, the server combines the results of the generative AI model and the sentiment analysis engine to evaluate the overall safety of the call. If the evaluation result is "safe," the server connects the call to the user's device. On the other hand, if the evaluation result is "risky," the server notifies the fraud department and blocks the call.

[0488] Specific examples

[0489] For example, if an elderly person receives a call from someone claiming to be from a bank, the system operates as follows:

[0490] 1. When a user receives a call, the device captures the voice data and sends it to the server.

[0491] 2. The server sends the audio data to the Google Cloud Speech-to-Text API and converts it into text data.

[0492] 3. The converted text data is summarized using OpenAI GPT-3. For example, a message such as "We have recently observed unusual transactions from your account. Please provide your personal information to confirm" is summarized as "Confirming unusual transactions on your account, requesting the provision of personal information."

[0493] 4. The server analyzes this summary text using OpenAI GPT-3 and rates it as "dangerous."

[0494] 5. At the same time, IBM Watson Tone Analyzer is used to analyze the user's emotions and assess whether the user is confused or nervous.

[0495] 6. The server combines these evaluation results and, if it determines the call is "risky," it reports it to the fraud department and blocks the call.

[0496] Prompt Sentence Examples

[0497] "If you receive a call from someone claiming to be a bank employee and they ask for personal information, transcribe the call and then summarize it to assess its security."

[0498] The present invention allows for safe and reliable call management by effectively assessing the risk of telephone fraud and taking into account the emotional state of the user.

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

[0500] Step 1:

[0501] When a user answers a call and the device starts a call, the server captures and receives the voice data in real time. Specifically, the device's microphone captures the user's voice and sends the voice data to the server using UDP or TCP protocols. The server then stores the received voice data in storage (e.g., Amazon S3).

[0502] Input: User's voice data

[0503] Output: Audio file saved to storage

[0504] Step 2:

[0505] The server retrieves the voice data from the storage and converts it into text data using voice recognition technology. It then activates a voice recognition engine (e.g., Google Cloud Speech-to-Text API) and sends the voice data to a cloud service. The converted text data is then stored in a database (e.g., MySQL).

[0506] Input: Audio file

[0507] Output: Character data stored in the database

[0508] Step 3:

[0509] The server retrieves text data from the database and condenses it using summarization technology. It then launches a summarization engine (e.g., OpenAI GPT-3) and sends the text data along with a prompt to the summarization engine. The summarized text data is then stored in the database.

[0510] Input: Character data

[0511] Output: Summary text stored in the database

[0512] Step 4:

[0513] The server analyzes the summarized text using a generative AI model to assess the risk of fraud and misconduct. The summarized text, along with a prompt, is sent to the generative AI (e.g., OpenAI GPT-3) to obtain a risk assessment result, which is also stored in a database.

[0514] Input: Summary text

[0515] Output: Risk assessment results stored in a database

[0516] Step 5:

[0517] The server passes the voice data to an emotion analysis engine to analyze the user's emotion. It then starts the emotion analysis engine (e.g., IBM Watson Tone Analyzer) and sends the voice data. The analysis results are stored in a database.

[0518] Input: Audio file

[0519] Output: Sentiment analysis results stored in a database

[0520] Step 6:

[0521] The server combines the evaluation results of the generative AI model and the sentiment analysis engine to perform a comprehensive safety assessment of the call content. The evaluation results are input into the integrated algorithm to perform a final safety assessment. If the evaluation result is judged to be "safe," the server instructs the call to be connected to the user's device. If the evaluation result is judged to be "risky," the server reports the call to the fraud department and blocks the call.

[0522] Input: Risk assessment results, sentiment analysis results

[0523] Output: Connect or block the call and report it to the fraud department

[0524] (Application example 2)

[0525] 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."

[0526] In modern society, telephone fraud is on the rise. The elderly and general public are especially at risk of falling prey to fraudulent calls, and traditional manual countermeasures have their limitations. There is also a need for systems that can automatically analyze the content of phone calls, evaluate their safety, and provide feedback and alerts to users.

[0527] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for recording voice, means for converting the recorded voice into text, means for summarizing the converted text, means for analyzing the summarized text using a generation AI and an emotion engine to evaluate the safety of communication and the user's emotions, means for controlling call connection based on the evaluation results, and means for feeding back the evaluation results to the user. This makes it possible to evaluate the safety of calls in real time, provide quick feedback to the user, and reduce the risk of fraud.

[0528] An "audio recording means" is a device or system that records telephone conversations in real time and stores them as digital audio data for subsequent processing.

[0529] A "means for converting recorded voice into text" is a device or system that uses voice recognition technology to convert recorded voice data into text data.

[0530] The "means for summarizing converted characters" is a device or system that analyzes text data, extracts important information, and summarizes it concisely.

[0531] "Means for analyzing using generative AI and an emotion engine to assess the safety of communication and the user's emotions" refers to a device or system that uses a generative AI model and an emotion analysis engine to assess emotions from the summarized text content and voice, and determine the safety of the call and the user's emotional state.

[0532] "Means for controlling call connections based on evaluation results" refers to a device or system that connects a call if it is judged to be safe based on the evaluation results of the generative AI and emotion engine, and blocks the connection if it is judged to be dangerous.

[0533] "Means for providing feedback of evaluation results to the user" refers to a device or system that notifies the user of the evaluation results by the generative AI and emotion engine, and provides information on the safety of the call and the connection status.

[0534] A "means for reporting to the fraud department" is a device or system that, if a call is deemed dangerous, notifies the fraud department of that information and prompts them to take appropriate action.

[0535] To implement this invention, the following system configuration and programs are required: The entire system is built around three elements: a server, a terminal, and a user.

[0536] System Configuration

[0537] 1. Server

[0538] As a means of recording audio, the server captures audio data during a call in real time and stores it as digital data.

[0539] To convert the recorded voice into text, the server uses voice recognition technology to convert the acquired voice data into text data. Specifically, it uses the speech_recognition library.

[0540] To summarize the converted text, the server uses natural language processing techniques to extract and summarize important information from the text data, for example, by using spacy to analyze the content of the text.

[0541] The summarized text is analyzed by the generative AI and emotion engine, and the generative AI model and emotion engine are activated as a means of evaluating the safety of communication and the user's emotions. The generative AI model uses a model from the transformers library, and the emotion engine uses the emotion_recognition library.

[0542] As a means of controlling call connections based on the evaluation results, the server summarizes the evaluation results from the AI ​​and emotion engine, and connects the call if it is judged to be safe, and stops the connection if it is judged to be dangerous.

[0543] As a means of providing feedback on the evaluation results to the user, the evaluation results are sent to the terminal and notified to the user.

[0544] 2. Terminal

[0545] The device receives the evaluation results sent from the server and provides visual or auditory feedback to the user, for example by displaying a warning message using the smartphone's notification function.

[0546] 3. Users

[0547] During a telephone call, the user receives feedback from the terminal and decides whether to continue or terminate the call as needed.

[0548] Specific examples

[0549] For example, if a user receives a call from someone claiming to be from a "bank," the call is first recorded by the server. The recorded voice data is converted into text data and recognized as, "We have recently observed unusual activity on your account. Please provide your personal information for verification." The server summarizes this text and summarizes it as, "Abnormal activity on account, request for personal information." This summarized text is then analyzed using a generative AI model and an emotion engine, and is rated as "dangerous." This evaluation result is sent to the device and displayed as a warning to the user. As a result, the user recognizes that the call may be fraudulent and can take appropriate action.

[0550] Prompt Sentence Examples

[0551] The following prompts are fed into the generative AI model for evaluation:

[0552] "Please summarize the following call and assess its likelihood of fraud or misconduct.\nCall content: {Call content}"

[0553] Thus, the present invention provides an effective system for reducing the risk of telephone fraud and keeping users safe.

[0554] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0555] Step 1:

[0556] When the server detects that the user has started a phone call, it starts recording the audio in real time by capturing the audio data using a microphone and storing the digital data on the server.

[0557] Input: User's call audio

[0558] Output: Digital audio data

[0559] Step 2:

[0560] The server converts the recorded voice data into text data using speech recognition technology. Specifically, it uses the speech_recognition library to convert voice to text.

[0561] Input: Digital audio data

[0562] Output: Text data

[0563] Step 3:

[0564] The server summarizes the text data using natural language processing techniques, such as the spacy library, to extract and summarize key information concisely.

[0565] Input: Text data

[0566] Output: Summarized text data

[0567] Step 4:

[0568] The server runs a generative AI model and an emotion engine to analyze the content of the summarized text and assess the safety of the call and the user's emotions. The generative AI model uses the transformers library, and the emotion engine uses the emotion_recognition library.

[0569] Input: Summarized text data

[0570] Output: Safety evaluation results and emotion evaluation results

[0571] Step 5:

[0572] The server controls the call connection based on the evaluation results of the generative AI and emotion engine. If the call is judged to be safe, the connection is maintained, and if it is judged to be dangerous, the connection is cut off.

[0573] Input: Safety evaluation results and emotion evaluation results

[0574] Output: Call connection status (connected / disconnected)

[0575] Step 6:

[0576] The server sends the evaluation results to the device and notifies the user, so the user can immediately know whether the call is safe or dangerous. The device displays this using its notification function.

[0577] Input: Safety evaluation results and emotion evaluation results

[0578] Output: User notification

[0579] Step 7:

[0580] The user receives feedback from the device and decides whether to continue or terminate the call, and if necessary, terminates the call and reports the call to the fraud department.

[0581] Input: User notification

[0582] Output: User behavior (continue / cancel call, report)

[0583] 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.

[0584] 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.

[0585] 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.

[0586] [Third embodiment]

[0587] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0588] 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.

[0589] 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).

[0590] 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.

[0591] 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.

[0592] 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).

[0593] 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.

[0594] 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.

[0595] 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.

[0596] 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.

[0597] 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.

[0598] 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."

[0599] The present invention provides a system that records the contents of phone calls, converts the voice data into text, then summarizes it, and evaluates the safety of the call using a generation AI. Below, we will explain the program processing and specific examples of this system.

[0600] System configuration and processing

[0601] 1. Call logging

[0602] When the user receives a call, the server uses a call recording means to capture the user's voice through a microphone, thereby obtaining the contents of the call as audio data.

[0603] 2. Transcription of audio data

[0604] The server uses the acquired voice data to activate a means for converting voice to text, which is performed using voice recognition technology to obtain the contents of the call as text data.

[0605] 3. Summary of Contents

[0606] The server then uses a means for summarizing the converted text data to concisely summarize the contents of the text, allowing the user to obtain the necessary information in a condensed form.

[0607] 4. AI-powered filtering

[0608] The server analyzes the summarized text and uses generative AI to assess the safety of the call, identifying whether the call is related to fraud or other fraudulent activity.

[0609] 5. Call connection control

[0610] The server activates a means to control the call connection based on the AI ​​evaluation results. Only if the evaluation result is judged to be "safe" will the call be connected to the user's device.

[0611] If the evaluation result is judged to be "risky," the server will not connect the call to the user's device and will report it to the fraud department.

[0612] Specific examples

[0613] For example, consider the case where an elderly person receives a phone call from someone claiming to be from their bank. The call is first recorded by a server. The recorded voice data is then converted into text. The converted text might say something like, "We've recently observed unusual transactions from your account. Please provide your personal information to verify."

[0614] The server summarizes this text into a short summary such as "Abnormal account transaction confirmation, request for personal information." The summarized text is then evaluated by a generative AI, which in this case determines it to be "risky."

[0615] As a result, the server will not connect the call to the user's device but will instead report it to the fraud department. This system significantly reduces the risk of seniors becoming victims of telephone fraud.

[0616] The processing flow will be explained below.

[0617] Step 1:

[0618] The user answers the call. The server activates the call recording function and captures the user's voice in real time through the microphone. This voice data is stored as is on the server.

[0619] Step 2:

[0620] The server analyzes the recorded voice data and activates a voice-to-text converter. Specifically, it uses voice recognition technology to convert the voice data into text data. The result of this conversion is a text representation of the contents of the call.

[0621] Step 3:

[0622] The server uses a means to summarize the converted text data, turning long conversations into concise summaries. The summarization process extracts important information and summarizes it in an easy-to-understand format.

[0623] Step 4:

[0624] The server then invokes the generative AI to analyze the summarized text. The AI ​​uses natural language processing techniques to analyze the text and assess whether the call is related to fraud or other misconduct, using historical data and pattern recognition.

[0625] Step 5:

[0626] The server receives the evaluation result from the AI ​​generator and activates a means to control the call connection based on the result. If the evaluation result is judged to be "safe," the server connects the call to the user's device.

[0627] Step 6:

[0628] If the evaluation results in a "risk" rating, the server will not connect the call to the user's device and will instead report it to the fraud department. This report will be sent along with the specific content of the call, allowing for a prompt response.

[0629] Example 1

[0630] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0631] In the past, elderly people and users with low security awareness have been at high risk of falling victim to telephone fraud and fraudulent calls. To solve this problem and ensure the safety of calls, a system is needed that can carefully examine the content of calls and determine their safety in real time.

[0632] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0633] In this invention, the server includes a means for recording calls, a means for converting the recorded call voice into text, a means for summarizing the converted text, a means for analyzing the summarized text using a generative AI model and evaluating the safety of the call content, and a means for controlling the call connection based on the evaluation result. This makes it possible to reduce the risk of elderly people and users with low security awareness becoming victims of telephone fraud and fraudulent calls.

[0634] "Call" refers to the act of communicating for the purpose of voice communication.

[0635] "Recording" refers to the act or means of saving audio data.

[0636] "Convert to text" refers to the process of converting voice data into text data.

[0637] "Summarization" refers to the process of shortening long text data, extracting important information, and summarizing it concisely.

[0638] A "generative AI model" refers to an algorithm or program that uses artificial intelligence to generate and analyze data.

[0639] "Analysis" refers to the process of examining data and understanding its properties and patterns.

[0640] "Evaluating call content for safety" refers to the evaluation process to determine whether the content of a call is related to fraud or fraudulent activity.

[0641] "Controlling call connections" refers to the process of allowing or denying call connections based on specific evaluation results.

[0642] "Surveillance and Response Unit" refers to a unit dedicated to responding to fraud and other misconduct when it occurs.

[0643] This invention provides a system that records the contents of phone calls, converts the voice data into text, then summarizes it and evaluates the safety of the call using generative AI. When a user answers a call, the server processes the call contents in real time, providing a safe calling environment.

[0644] 1. Call logging

[0645] When a user answers a call, the server automatically starts recording the call. Specifically, the server captures audio data in real time through the device's microphone and saves it as audio data. The saved audio file format is typically WAV or MP3.

[0646] 2. Transcription of audio data

[0647] The server uses speech recognition software such as Google Cloud Speech-to-Text API to convert the recorded voice data into text data. The server sends the audio file to the API, which processes it and obtains the text data. The obtained text data is a direct transcription of the contents of the call.

[0648] 3. Summary of Contents

[0649] The server uses natural language processing techniques such as the Transformers library to summarize the retrieved text data. The text data is input into a summarization model to obtain a summary sentence. For example, the text data "Hello, this is a message from the bank" is converted into a summary sentence such as "A message from the bank."

[0650] 4. AI-powered filtering

[0651] The server uses a generative AI model (e.g., GPT-4) to evaluate the safety of the call content based on the summary. It generates a prompt for the generative AI model and obtains the evaluation result by inputting the summary into the prompt. An example of a prompt is, "Please evaluate the following call content. Summary: 'Confirmation of abnormal account transactions, request for personal information'. Is this call safe or dangerous?"

[0652] 5. Call connection control

[0653] The server manages the call connection based on the evaluation results of the generated AI model. The call is connected to the user's device only if the evaluation result is judged to be "safe." If the result is judged to be "dangerous," the call connection is cut off and a report is sent to the monitoring department. This process ensures the safety of the user.

[0654] Specific examples

[0655] For example, if an elderly person receives a call from someone claiming to be from their bank, the call is first recorded by the server. The recorded voice data is converted into text data using the Google Cloud Speech-to-Text API. The converted text may then say something like, "We have recently observed unusual transactions in your account. Please provide your personal information to confirm." This text data is then summarized using the Transformers library as "Confirming unusual account transactions, requesting personal information."

[0656] The summarized text is fed into the GPT-4 model, and the generative AI model determines that the call is "risky." As a result, the server does not connect the call to the user's device and reports it to the surveillance department. This significantly reduces the risk of seniors becoming victims of telephone fraud.

[0657] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0658] Step 1:

[0659] When a user answers a call, the server starts recording the call. Specifically, the server captures audio data in real time from the device's microphone and saves it as an audio file, typically in WAV or MP3 format.

[0660] Input: User's call audio

[0661] Output: Audio file (WAV or MP3 format)

[0662] Step 2:

[0663] The server uses speech recognition software (e.g., Google Cloud Speech-to-Text API) to convert the recorded audio file into text. The server sends the audio file to the API, which processes it and obtains the text data.

[0664] Input: Audio file

[0665] Output: Text data

[0666] Step 3:

[0667] The server uses natural language processing techniques (e.g., the Transformers library) to summarize the acquired text data. The summarization model narrows down the input text data to the important information, shortens it, and generates a summary sentence.

[0668] Input: Text data

[0669] Output: Summary

[0670] Step 4:

[0671] The server evaluates the safety of the call content using a generative AI model (e.g., GPT-4) based on the summary. Specifically, the server generates a prompt for the generative AI model, inputs the summary along with this prompt, and obtains the evaluation result. An example of a prompt is, "Please evaluate the following call content. Summary: 'Confirmation of abnormal account transactions, request for personal information'. Is this call safe or dangerous?"

[0672] Input: Summary, prompt

[0673] Output: Evaluation result (safe / dangerous)

[0674] Step 5:

[0675] The server controls the call connection based on the evaluation results obtained by the generative AI model. The call is connected to the user's device only if the evaluation result is judged to be "safe." If the evaluation result is judged to be "dangerous," the call connection is cut off and a notification is sent to the monitoring and response department. Specifically, the server controls the communication software according to the evaluation result and issues a warning to the security department if necessary.

[0676] Input: Evaluation result

[0677] Output: Connect or disconnect the call, notify the monitoring department

[0678] (Application example 1)

[0679] 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."

[0680] In recent years, telephone fraud and malicious communications targeting the elderly have been increasing, creating a need to protect users from such fraudulent activities. However, current call monitoring systems have difficulty detecting fraud in real time, and there is also concern that legitimate calls may be blocked due to false positives. Therefore, it is necessary to develop a system that can accurately and quickly evaluate the security of calls and detect fraudulent calls in real time.

[0681] 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.

[0682] In this invention, the server includes a means for recording voice, a means for converting voice to text, a means for summarizing the text, a means for analyzing the summarized text using a generative AI model and evaluating the safety of communication, a means for controlling call connections based on the evaluation result, and a means for reporting to an anti-fraud department if the evaluation is unsafe. This makes it possible to evaluate the safety of calls quickly and with high accuracy, and to detect and deal with fraudulent calls in real time.

[0683] "Audio recording means" is a mechanism for collecting audio data during a call and storing it in a digital format.

[0684] "Means for converting voice to text" refers to technology that analyzes recorded voice data and converts its contents into text data.

[0685] The "text summarization means" is an algorithm that concisely summarizes the converted text data and extracts only the necessary information.

[0686] A "generative AI model" is a trained model that uses artificial intelligence to analyze data and make inferences.

[0687] "Means for assessing the security of communications" refers to the process of using generative AI models to determine whether a call contains fraud or scam content.

[0688] The "means for controlling call connection" is a function that determines whether to connect a call to a user terminal or block the call based on the evaluation result.

[0689] "Means to report to fraud prevention department" is a function that notifies a specialized department if the evaluated content of a call is determined to be fraudulent.

[0690] The present invention relates to a system for monitoring telephone call content in real time to detect fraud and other illegal activities. Specific embodiments of this system will be described below.

[0691] First, when a user answers a call, the server uses the device's microphone to record the call audio. This audio data is then saved in a digital format. Next, speech recognition technology is used to convert the audio data into text. Specifically, this technology uses Google's speech recognition API.

[0692] The converted text data is then further summarized. For summarization, we use the Hugging Face transformers library. As a concrete example, we use the T5 model to concisely summarize the text content.

[0693] The next step is to use a generative AI model to assess the safety of the summarized text. This assessment is performed using a custom SafetyEvaluator model. The generative AI model is fed a specific prompt, such as "Please assess whether this call is fraud-related. Provide the following summarized text: 'Confirm unusual account activity, request for personal information.'"

[0694] After the evaluation result is obtained, the call connection is controlled based on the result. If the evaluation result is determined to be "safe," the call is successfully connected to the user's device. Conversely, if the evaluation result is "unsafe," the call is blocked and automatically reported to the fraud department. The report contains the necessary information to enable the specialized department to take action quickly.

[0695] In this way, the present invention makes it possible to accurately and quickly evaluate the security of calls and detect and deal with fraudulent calls in real time, thereby protecting users from fraudulent activities and significantly reducing the damage caused by telephone fraud and other crimes.

[0696] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0697] Step 1:

[0698] When a user receives a call, the device's microphone is used to record the call audio. The call audio data is transmitted from the microphone to a server in real time and stored in a digital format. The input is raw audio data, and the output is digital audio data.

[0699] Step 2:

[0700] The server converts the recorded voice data into text data using speech recognition technology. This process uses Google's speech recognition API. The input is digital voice data and the output is text data. The server analyzes the voice data and converts each voice segment into the corresponding text.

[0701] Step 3:

[0702] The server summarizes the converted text data. This summarization utilizes Hugging Face's transformers library, specifically the T5 model. The input is the converted text data, and the output is the summary text. The server analyzes the text data, extracts important information, and summarizes it concisely.

[0703] Step 4:

[0704] The server analyzes the summarized text data and evaluates safety using a generative AI model. Using a custom SafetyEvaluator model, the input is the summarized text and the output is the safety evaluation result. For example, this process inputs a prompt such as, "Please evaluate whether this call is fraud-related. Provide the following summarized text: 'Confirm abnormal account transactions, request for personal information.'"

[0705] Step 5:

[0706] The server controls the call connection based on the evaluation results. If the evaluation result is judged to be "safe," the call is connected to the user's device. The input is the safety evaluation result, and the output is whether the call can be connected. Conversely, if the call is judged to be "unsafe," the call is blocked and an automatic report is sent to the fraud prevention department. The report includes a summary of the specific call content and details of the evaluation result.

[0707] These processing steps enable accurate and rapid assessment of call security and real-time detection and response to fraudulent calls.

[0708] 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.

[0709] The present invention provides a system that records the contents of phone calls, converts the voice data into text, then summarizes it, and evaluates the safety of the call using a generative AI and emotion engine. Below, we will explain the program processing and specific examples of this system.

[0710] System configuration and processing

[0711] 1. Call logging

[0712] When the user answers the phone, the server uses a call recording method to capture the user's voice through a microphone, and this voice data is stored on the server as is.

[0713] 2. Transcription of audio data

[0714] The server uses the acquired voice data to activate a means for converting voice to text, which is performed using voice recognition technology to obtain the contents of the call as text data.

[0715] 3. Summary of Contents

[0716] The server then uses a means for summarizing the converted text data to concisely summarize the contents of the text, allowing the user to obtain the necessary information in a condensed form.

[0717] 4. AI-powered filtering

[0718] The server then invokes the generative AI to analyze the summarized text. The AI ​​uses natural language processing techniques to analyze the text and assess whether the call is related to fraud or other misconduct, using historical data and pattern recognition.

[0719] 5. Emotion Recognition by Emotion Engine

[0720] The server uses the user's voice data to activate the emotion engine, which uses voice analysis technology to assess the user's emotions and stress level, thereby detecting emotions such as confusion or nervousness.

[0721] 6. Integration of evaluation results and call connection control

[0722] The server combines the evaluation results of the generative AI and the emotion engine to perform a comprehensive safety assessment of the call content. If the assessment result is deemed "safe," the server connects the call to the user's device.

[0723] If the evaluation result is judged to be "risky," the server will not connect the call to the user's device and will report it to the fraud department.

[0724] Specific examples

[0725] For example, consider the case where an elderly person receives a phone call from someone claiming to be from a bank. The call is first recorded by a server. The recorded voice data is then converted into text. The converted text might say something like, "We've recently observed unusual transactions from your account. Please provide your personal information to verify."

[0726] The server summarizes this text into a short summary such as "Abnormal account transaction confirmation, request for personal information." The summarized text is then evaluated by a generative AI, which in this case determines it to be "risky."

[0727] At the same time, the emotion engine analyzes the voice data and determines whether the user is confused or nervous. This information is also added to the overall evaluation. As a result, the server will not connect the call to the user's device and will report it to the fraud department.

[0728] This system can significantly reduce the risk of elderly people becoming victims of telephone fraud. In addition, by taking into account the user's emotions, it can provide a more accurate evaluation.

[0729] The processing flow will be explained below.

[0730] Step 1:

[0731] The user answers the call. The server activates the call recording function and captures the user's voice in real time through the microphone. This voice data is stored as is on the server.

[0732] Step 2:

[0733] The server uses the recorded voice data to activate a voice-to-text conversion means. Specifically, it uses voice recognition technology to convert the voice data into text data. The result of this conversion is a text representation of the contents of the call.

[0734] Step 3:

[0735] The server uses a means to summarize the converted text data, turning long conversations into concise summaries. The summarization process extracts important information and summarizes it in an easy-to-understand format.

[0736] Step 4:

[0737] The server then invokes the generative AI to analyze the summarized text. The AI ​​uses natural language processing techniques to analyze the text and assess whether the call is related to fraud or other misconduct, using historical data and pattern recognition.

[0738] Step 5:

[0739] The server receives the evaluation results from the AI ​​generator, along with the user's voice data, which then activates the emotion engine. The emotion engine uses voice analysis technology to evaluate the user's emotions and stress level, detecting whether the user is confused, nervous, or relieved.

[0740] Step 6:

[0741] The server combines the results of the generative AI and the emotion engine to perform a comprehensive safety assessment of the call content, specifically determining whether the likelihood of fraud is high or low, and whether the user is nervous.

[0742] Step 7:

[0743] If the evaluation result is deemed "safe," the server connects the call to the user's device, allowing the user to continue the call with peace of mind.

[0744] Step 8:

[0745] If the evaluation results in a "risk" rating, the server will not connect the call to the user's device and will instead report it to the fraud department. This report will include the specific content of the call and information about the user's emotional state, allowing for a prompt response.

[0746] Specific examples

[0747] For example, consider the case where an elderly person receives a phone call from someone claiming to be from their bank. The call is first recorded by a server. The recorded voice data is then converted into text. The converted text might say something like, "We've recently observed unusual transactions from your account. Please provide your personal information to verify."

[0748] The server summarizes this text into a short summary such as "Abnormal account transaction confirmation, request for personal information." The summarized text is then evaluated by a generative AI, which in this case determines it to be "risky."

[0749] At the same time, the emotion engine analyzes the voice data and evaluates whether the user is confused or nervous. This information is also taken into account in the overall evaluation. As a result, the server reports the call to the fraud department without connecting it to the user's device. This system can significantly reduce the risk of elderly people becoming victims of telephone fraud. Furthermore, taking the user's emotions into account allows for more accurate evaluation.

[0750] Example 2

[0751] 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."

[0752] In recent years, telephone fraud targeting the elderly has been increasing, and an effective system to prevent such fraud is required. To prevent such fraud, it is necessary to analyze the contents of calls quickly and accurately and to identify fraudulent or dangerous calls in real time. Conventional systems do not take into account the user's emotions or stress state, making it difficult to improve accuracy.

[0753] 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.

[0754] In this invention, the server includes means for recording voice, means for converting the recorded voice into text, means for summarizing the converted text, means for analyzing the summarized text using a generative AI model to assess the risk of fraud or misconduct, means for analyzing the recorded voice using an emotion analysis engine to assess the emotional state, means for integrating the evaluation results of the generative AI model and the emotion analysis engine to comprehensively assess the safety of communication, and means for controlling call connections based on the evaluation results. This makes it possible to assess the risk of telephone fraud with high accuracy and comprehensively determine the safety of calls while taking the user's emotional state into consideration.

[0755] "Audio recording means" refers to devices or techniques used to capture the contents of telephone conversations.

[0756] The "means for converting recorded voice into text" refers to a voice recognition technology for converting voice data into text data.

[0757] The "means for summarizing converted characters" is a technique for concisely summarizing text data that has been recognized by speech recognition.

[0758] "Means for analyzing using a generative AI model to assess the risk of fraud or misconduct" refers to a technology that uses AI technology to analyze text data and assess the possibility of fraud or misconduct.

[0759] The "means for analyzing using an emotion analysis engine and assessing the emotional state" is a technology that analyzes voice data to assess the user's emotions and stress state.

[0760] "Means for integrating evaluation results and comprehensively assessing the safety of communication" refers to a technology that integrates the evaluation results obtained from a generative AI model and an emotion analysis engine to comprehensively judge the safety of a call.

[0761] The "means for controlling call connection based on the evaluation result" is a technology for controlling call connection based on the safety evaluation result of the call content.

[0762] This invention relates to a system that records phone conversations, converts the audio data into text, then summarizes it and evaluates the safety of the call using generative AI and an emotion engine.

[0763] First, when a user answers a phone call, the contents of the call are recorded by the device. The recorded voice data is immediately sent to the server and stored on the server side. Specifically, the user's voice is picked up through the device's microphone, and the voice data is transferred to the server using a network protocol (e.g., UDP or TCP). The server then stores this data in cloud storage (e.g., Amazon S3).

[0764] The server then converts the recorded audio data into text data using speech recognition technology, specifically a speech recognition engine (e.g., Google Cloud Speech-to-Text API). The server sends the audio file to the API and stores the returned text data in a database (e.g., MySQL).

[0765] The server then uses a summarization engine (e.g., OpenAI GPT-3) to summarize the converted text data. The server sends the text data along with the prompt to the summarization engine and receives the summarized text data. The summary results are also stored in a database.

[0766] The server then uses a generative AI model to perform a risk assessment on the summarized text. Specifically, the summarized text is sent to a generative AI (e.g., OpenAI GPT-3) again to assess the likelihood of fraud or misconduct. The results of this assessment are also stored on the server.

[0767] Furthermore, the server launches an emotion analysis engine (e.g., IBM Watson Tone Analyzer) to recognize and evaluate the user's emotions from the recorded voice data. The server sends the voice data to the emotion analysis engine and receives the analysis results. These evaluation results are also stored in the database.

[0768] Finally, the server combines the results of the generative AI model and the sentiment analysis engine to evaluate the overall safety of the call. If the evaluation result is "safe," the server connects the call to the user's device. On the other hand, if the evaluation result is "risky," the server notifies the fraud department and blocks the call.

[0769] Specific examples

[0770] For example, if an elderly person receives a call from someone claiming to be from a bank, the system operates as follows:

[0771] 1. When a user receives a call, the device captures the voice data and sends it to the server.

[0772] 2. The server sends the audio data to the Google Cloud Speech-to-Text API and converts it into text data.

[0773] 3. The converted text data is summarized using OpenAI GPT-3. For example, a message such as "We have recently observed unusual transactions from your account. Please provide your personal information to confirm" is summarized as "Confirming unusual transactions on your account, requesting the provision of personal information."

[0774] 4. The server analyzes this summary text using OpenAI GPT-3 and rates it as "dangerous."

[0775] 5. At the same time, IBM Watson Tone Analyzer is used to analyze the user's emotions and assess whether the user is confused or nervous.

[0776] 6. The server combines these evaluation results and, if it determines the call is "risky," it reports it to the fraud department and blocks the call.

[0777] Prompt Sentence Examples

[0778] "If you receive a call from someone claiming to be a bank employee and they ask for personal information, transcribe the call and then summarize it to assess its security."

[0779] The present invention allows for safe and reliable call management by effectively assessing the risk of telephone fraud and taking into account the emotional state of the user.

[0780] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0781] Step 1:

[0782] When a user answers a call and the device starts a call, the server captures and receives the voice data in real time. Specifically, the device's microphone captures the user's voice and sends the voice data to the server using UDP or TCP protocols. The server then stores the received voice data in storage (e.g., Amazon S3).

[0783] Input: User's voice data

[0784] Output: Audio file saved to storage

[0785] Step 2:

[0786] The server retrieves the voice data from the storage and converts it into text data using voice recognition technology. It then activates a voice recognition engine (e.g., Google Cloud Speech-to-Text API) and sends the voice data to a cloud service. The converted text data is then stored in a database (e.g., MySQL).

[0787] Input: Audio file

[0788] Output: Character data stored in the database

[0789] Step 3:

[0790] The server retrieves text data from the database and condenses it using summarization technology. It then launches a summarization engine (e.g., OpenAI GPT-3) and sends the text data along with a prompt to the summarization engine. The summarized text data is then stored in the database.

[0791] Input: Character data

[0792] Output: Summary text stored in the database

[0793] Step 4:

[0794] The server analyzes the summarized text using a generative AI model to assess the risk of fraud and misconduct. The summarized text, along with a prompt, is sent to the generative AI (e.g., OpenAI GPT-3) to obtain a risk assessment result, which is also stored in a database.

[0795] Input: Summary text

[0796] Output: Risk assessment results stored in a database

[0797] Step 5:

[0798] The server passes the voice data to an emotion analysis engine to analyze the user's emotion. It then starts the emotion analysis engine (e.g., IBM Watson Tone Analyzer) and sends the voice data. The analysis results are stored in a database.

[0799] Input: Audio file

[0800] Output: Sentiment analysis results stored in a database

[0801] Step 6:

[0802] The server combines the evaluation results of the generative AI model and the sentiment analysis engine to perform a comprehensive safety assessment of the call content. The evaluation results are input into the integrated algorithm to perform a final safety assessment. If the evaluation result is judged to be "safe," the server instructs the call to be connected to the user's device. If the evaluation result is judged to be "risky," the server reports the call to the fraud department and blocks the call.

[0803] Input: Risk assessment results, sentiment analysis results

[0804] Output: Connect or block the call and report it to the fraud department

[0805] (Application example 2)

[0806] 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."

[0807] In modern society, telephone fraud is on the rise. The elderly and general public are especially at risk of falling prey to fraudulent calls, and traditional manual countermeasures have their limitations. There is also a need for systems that can automatically analyze the content of phone calls, evaluate their safety, and provide feedback and alerts to users.

[0808] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for recording voice, means for converting the recorded voice into text, means for summarizing the converted text, means for analyzing the summarized text using a generation AI and an emotion engine to evaluate the safety of communication and the user's emotions, means for controlling call connection based on the evaluation results, and means for feeding back the evaluation results to the user. This makes it possible to evaluate the safety of calls in real time, provide quick feedback to the user, and reduce the risk of fraud.

[0809] An "audio recording means" is a device or system that records telephone conversations in real time and stores them as digital audio data for subsequent processing.

[0810] A "means for converting recorded voice into text" is a device or system that uses voice recognition technology to convert recorded voice data into text data.

[0811] The "means for summarizing converted characters" is a device or system that analyzes text data, extracts important information, and summarizes it concisely.

[0812] "Means for analyzing using generative AI and an emotion engine to assess the safety of communication and the user's emotions" refers to a device or system that uses a generative AI model and an emotion analysis engine to assess emotions from the summarized text content and voice, and determine the safety of the call and the user's emotional state.

[0813] "Means for controlling call connections based on evaluation results" refers to a device or system that connects a call if it is judged to be safe based on the evaluation results of the generative AI and emotion engine, and blocks the connection if it is judged to be dangerous.

[0814] "Means for providing feedback of evaluation results to the user" refers to a device or system that notifies the user of the evaluation results by the generative AI and emotion engine, and provides information on the safety of the call and the connection status.

[0815] A "means for reporting to the fraud department" is a device or system that, if a call is deemed dangerous, notifies the fraud department of that information and prompts them to take appropriate action.

[0816] To implement this invention, the following system configuration and programs are required: The entire system is built around three elements: a server, a terminal, and a user.

[0817] System Configuration

[0818] 1. Server

[0819] As a means of recording audio, the server captures audio data during a call in real time and stores it as digital data.

[0820] To convert the recorded voice into text, the server uses voice recognition technology to convert the acquired voice data into text data. Specifically, it uses the speech_recognition library.

[0821] To summarize the converted text, the server uses natural language processing techniques to extract and summarize important information from the text data, for example, by using spacy to analyze the content of the text.

[0822] The summarized text is analyzed by the generative AI and emotion engine, and the generative AI model and emotion engine are activated as a means of evaluating the safety of communication and the user's emotions. The generative AI model uses a model from the transformers library, and the emotion engine uses the emotion_recognition library.

[0823] As a means of controlling call connections based on the evaluation results, the server summarizes the evaluation results from the AI ​​and emotion engine, and connects the call if it is judged to be safe, and stops the connection if it is judged to be dangerous.

[0824] As a means of providing feedback on the evaluation results to the user, the evaluation results are sent to the terminal and notified to the user.

[0825] 2. Terminal

[0826] The device receives the evaluation results sent from the server and provides visual or auditory feedback to the user, for example by displaying a warning message using the smartphone's notification function.

[0827] 3. Users

[0828] During a telephone call, the user receives feedback from the terminal and decides whether to continue or terminate the call as needed.

[0829] Specific examples

[0830] For example, if a user receives a call from someone claiming to be from a "bank," the call is first recorded by the server. The recorded voice data is converted into text data and recognized as, "We have recently observed unusual activity on your account. Please provide your personal information for verification." The server summarizes this text and summarizes it as, "Abnormal activity on account, request for personal information." This summarized text is then analyzed using a generative AI model and an emotion engine, and is rated as "dangerous." This evaluation result is sent to the device and displayed as a warning to the user. As a result, the user recognizes that the call may be fraudulent and can take appropriate action.

[0831] Prompt Sentence Examples

[0832] The following prompts are fed into the generative AI model for evaluation:

[0833] "Please summarize the following call and assess its likelihood of fraud or misconduct.\nCall content: {Call content}"

[0834] Thus, the present invention provides an effective system for reducing the risk of telephone fraud and keeping users safe.

[0835] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0836] Step 1:

[0837] When the server detects that the user has started a phone call, it starts recording the audio in real time by capturing the audio data using a microphone and storing the digital data on the server.

[0838] Input: User's call audio

[0839] Output: Digital audio data

[0840] Step 2:

[0841] The server converts the recorded voice data into text data using speech recognition technology. Specifically, it uses the speech_recognition library to convert voice to text.

[0842] Input: Digital audio data

[0843] Output: Text data

[0844] Step 3:

[0845] The server summarizes the text data using natural language processing techniques, such as the spacy library, to extract and summarize key information concisely.

[0846] Input: Text data

[0847] Output: Summarized text data

[0848] Step 4:

[0849] The server runs a generative AI model and an emotion engine to analyze the content of the summarized text and assess the safety of the call and the user's emotions. The generative AI model uses the transformers library, and the emotion engine uses the emotion_recognition library.

[0850] Input: Summarized text data

[0851] Output: Safety evaluation results and emotion evaluation results

[0852] Step 5:

[0853] The server controls the call connection based on the evaluation results of the generative AI and emotion engine. If the call is judged to be safe, the connection is maintained, and if it is judged to be dangerous, the connection is cut off.

[0854] Input: Safety evaluation results and emotion evaluation results

[0855] Output: Call connection status (connected / disconnected)

[0856] Step 6:

[0857] The server sends the evaluation results to the device and notifies the user, so the user can immediately know whether the call is safe or dangerous. The device displays this using its notification function.

[0858] Input: Safety evaluation results and emotion evaluation results

[0859] Output: User notification

[0860] Step 7:

[0861] The user receives feedback from the device and decides whether to continue or terminate the call, and if necessary, terminates the call and reports the call to the fraud department.

[0862] Input: User notification

[0863] Output: User behavior (continue / cancel call, report)

[0864] 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.

[0865] 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.

[0866] 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.

[0867] [Fourth embodiment]

[0868] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[0869] 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.

[0870] 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).

[0871] 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.

[0872] 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.

[0873] 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).

[0874] 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.

[0875] 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.

[0876] 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.

[0877] 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.

[0878] 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.

[0879] 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.

[0880] 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."

[0881] The present invention provides a system that records the contents of phone calls, converts the voice data into text, then summarizes it, and evaluates the safety of the call using a generation AI. Below, we will explain the program processing and specific examples of this system.

[0882] System configuration and processing

[0883] 1. Call logging

[0884] When the user receives a call, the server uses a call recording means to capture the user's voice through a microphone, thereby obtaining the contents of the call as audio data.

[0885] 2. Transcription of audio data

[0886] The server uses the acquired voice data to activate a means for converting voice to text, which is performed using voice recognition technology to obtain the contents of the call as text data.

[0887] 3. Summary of Contents

[0888] The server then uses a means for summarizing the converted text data to concisely summarize the contents of the text, allowing the user to obtain the necessary information in a condensed form.

[0889] 4. AI-powered filtering

[0890] The server analyzes the summarized text and uses generative AI to assess the safety of the call, identifying whether the call is related to fraud or other fraudulent activity.

[0891] 5. Call connection control

[0892] The server activates a means to control the call connection based on the AI ​​evaluation results. Only if the evaluation result is judged to be "safe" will the call be connected to the user's device.

[0893] If the evaluation result is judged to be "risky," the server will not connect the call to the user's device and will report it to the fraud department.

[0894] Specific examples

[0895] For example, consider the case where an elderly person receives a phone call from someone claiming to be from their bank. The call is first recorded by a server. The recorded voice data is then converted into text. The converted text might say something like, "We've recently observed unusual transactions from your account. Please provide your personal information to verify."

[0896] The server summarizes this text into a short summary such as "Abnormal account transaction confirmation, request for personal information." The summarized text is then evaluated by a generative AI, which in this case determines it to be "risky."

[0897] As a result, the server will not connect the call to the user's device but will instead report it to the fraud department. This system significantly reduces the risk of seniors becoming victims of telephone fraud.

[0898] The processing flow will be explained below.

[0899] Step 1:

[0900] The user answers the call. The server activates the call recording function and captures the user's voice in real time through the microphone. This voice data is stored as is on the server.

[0901] Step 2:

[0902] The server analyzes the recorded voice data and activates a voice-to-text converter. Specifically, it uses voice recognition technology to convert the voice data into text data. The result of this conversion is a text representation of the contents of the call.

[0903] Step 3:

[0904] The server uses a means to summarize the converted text data, turning long conversations into concise summaries. The summarization process extracts important information and summarizes it in an easy-to-understand format.

[0905] Step 4:

[0906] The server then invokes the generative AI to analyze the summarized text. The AI ​​uses natural language processing techniques to analyze the text and assess whether the call is related to fraud or other misconduct, using historical data and pattern recognition.

[0907] Step 5:

[0908] The server receives the evaluation result from the AI ​​generator and activates a means to control the call connection based on the result. If the evaluation result is judged to be "safe," the server connects the call to the user's device.

[0909] Step 6:

[0910] If the evaluation results in a "risk" rating, the server will not connect the call to the user's device and will instead report it to the fraud department. This report will be sent along with the specific content of the call, allowing for a prompt response.

[0911] Example 1

[0912] 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."

[0913] In the past, elderly people and users with low security awareness have been at high risk of falling victim to telephone fraud and fraudulent calls. To solve this problem and ensure the safety of calls, a system is needed that can carefully examine the content of calls and determine their safety in real time.

[0914] 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.

[0915] In this invention, the server includes a means for recording calls, a means for converting the recorded call voice into text, a means for summarizing the converted text, a means for analyzing the summarized text using a generative AI model and evaluating the safety of the call content, and a means for controlling the call connection based on the evaluation result. This makes it possible to reduce the risk of elderly people and users with low security awareness becoming victims of telephone fraud and fraudulent calls.

[0916] "Call" refers to the act of communicating for the purpose of voice communication.

[0917] "Recording" refers to the act or means of saving audio data.

[0918] "Convert to text" refers to the process of converting voice data into text data.

[0919] "Summarization" refers to the process of shortening long text data, extracting important information, and summarizing it concisely.

[0920] A "generative AI model" refers to an algorithm or program that uses artificial intelligence to generate and analyze data.

[0921] "Analysis" refers to the process of examining data and understanding its properties and patterns.

[0922] "Evaluating call content for safety" refers to the evaluation process to determine whether the content of a call is related to fraud or fraudulent activity.

[0923] "Controlling call connections" refers to the process of allowing or denying call connections based on specific evaluation results.

[0924] "Surveillance and Response Unit" refers to a unit dedicated to responding to fraud and other misconduct when it occurs.

[0925] This invention provides a system that records the contents of phone calls, converts the voice data into text, then summarizes it and evaluates the safety of the call using generative AI. When a user answers a call, the server processes the call contents in real time, providing a safe calling environment.

[0926] 1. Call logging

[0927] When a user answers a call, the server automatically starts recording the call. Specifically, the server captures audio data in real time through the device's microphone and saves it as audio data. The saved audio file format is typically WAV or MP3.

[0928] 2. Transcription of audio data

[0929] The server uses speech recognition software such as Google Cloud Speech-to-Text API to convert the recorded voice data into text data. The server sends the audio file to the API, which processes it and obtains the text data. The obtained text data is a direct transcription of the contents of the call.

[0930] 3. Summary of Contents

[0931] The server uses natural language processing techniques such as the Transformers library to summarize the retrieved text data. The text data is input into a summarization model to obtain a summary sentence. For example, the text data "Hello, this is a message from the bank" is converted into a summary sentence such as "A message from the bank."

[0932] 4. AI-powered filtering

[0933] The server uses a generative AI model (e.g., GPT-4) to evaluate the safety of the call content based on the summary. It generates a prompt for the generative AI model and obtains the evaluation result by inputting the summary into the prompt. An example of a prompt is, "Please evaluate the following call content. Summary: 'Confirmation of abnormal account transactions, request for personal information'. Is this call safe or dangerous?"

[0934] 5. Call connection control

[0935] The server manages the call connection based on the evaluation results of the generated AI model. The call is connected to the user's device only if the evaluation result is judged to be "safe." If the result is judged to be "dangerous," the call connection is cut off and a report is sent to the monitoring department. This process ensures the safety of the user.

[0936] Specific examples

[0937] For example, if an elderly person receives a call from someone claiming to be from their bank, the call is first recorded by the server. The recorded voice data is converted into text data using the Google Cloud Speech-to-Text API. The converted text may then say something like, "We have recently observed unusual transactions in your account. Please provide your personal information to confirm." This text data is then summarized using the Transformers library as "Confirming unusual account transactions, requesting personal information."

[0938] The summarized text is fed into the GPT-4 model, and the generative AI model determines that the call is "risky." As a result, the server does not connect the call to the user's device and reports it to the surveillance department. This significantly reduces the risk of seniors becoming victims of telephone fraud.

[0939] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0940] Step 1:

[0941] When a user answers a call, the server starts recording the call. Specifically, the server captures audio data in real time from the device's microphone and saves it as an audio file, typically in WAV or MP3 format.

[0942] Input: User's call audio

[0943] Output: Audio file (WAV or MP3 format)

[0944] Step 2:

[0945] The server uses speech recognition software (e.g., Google Cloud Speech-to-Text API) to convert the recorded audio file into text. The server sends the audio file to the API, which processes it and obtains the text data.

[0946] Input: Audio file

[0947] Output: Text data

[0948] Step 3:

[0949] The server uses natural language processing techniques (e.g., the Transformers library) to summarize the acquired text data. The summarization model narrows down the input text data to the important information, shortens it, and generates a summary sentence.

[0950] Input: Text data

[0951] Output: Summary

[0952] Step 4:

[0953] The server evaluates the safety of the call content using a generative AI model (e.g., GPT-4) based on the summary. Specifically, the server generates a prompt for the generative AI model, inputs the summary along with this prompt, and obtains the evaluation result. An example of a prompt is, "Please evaluate the following call content. Summary: 'Confirmation of abnormal account transactions, request for personal information'. Is this call safe or dangerous?"

[0954] Input: Summary, prompt

[0955] Output: Evaluation result (safe / dangerous)

[0956] Step 5:

[0957] The server controls the call connection based on the evaluation results obtained by the generative AI model. The call is connected to the user's device only if the evaluation result is judged to be "safe." If the evaluation result is judged to be "dangerous," the call connection is cut off and a notification is sent to the monitoring and response department. Specifically, the server controls the communication software according to the evaluation result and issues a warning to the security department if necessary.

[0958] Input: Evaluation result

[0959] Output: Connect or disconnect the call, notify the monitoring department

[0960] (Application example 1)

[0961] 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."

[0962] In recent years, telephone fraud and malicious communications targeting the elderly have been increasing, creating a need to protect users from such fraudulent activities. However, current call monitoring systems have difficulty detecting fraud in real time, and there is also concern that legitimate calls may be blocked due to false positives. Therefore, it is necessary to develop a system that can accurately and quickly evaluate the security of calls and detect fraudulent calls in real time.

[0963] 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.

[0964] In this invention, the server includes a means for recording voice, a means for converting voice to text, a means for summarizing the text, a means for analyzing the summarized text using a generative AI model and evaluating the safety of communication, a means for controlling call connections based on the evaluation result, and a means for reporting to an anti-fraud department if the evaluation is unsafe. This makes it possible to evaluate the safety of calls quickly and with high accuracy, and to detect and deal with fraudulent calls in real time.

[0965] "Audio recording means" is a mechanism for collecting audio data during a call and storing it in a digital format.

[0966] "Means for converting voice to text" refers to technology that analyzes recorded voice data and converts its contents into text data.

[0967] The "text summarization means" is an algorithm that concisely summarizes the converted text data and extracts only the necessary information.

[0968] A "generative AI model" is a trained model that uses artificial intelligence to analyze data and make inferences.

[0969] "Means for assessing the security of communications" refers to the process of using generative AI models to determine whether a call contains fraud or scam content.

[0970] The "means for controlling call connection" is a function that determines whether to connect a call to a user terminal or block the call based on the evaluation result.

[0971] "Means to report to fraud prevention department" is a function that notifies a specialized department if the evaluated content of a call is determined to be fraudulent.

[0972] The present invention relates to a system for monitoring telephone call content in real time to detect fraud and other illegal activities. Specific embodiments of this system will be described below.

[0973] First, when a user answers a call, the server uses the device's microphone to record the call audio. This audio data is then saved in a digital format. Next, speech recognition technology is used to convert the audio data into text. Specifically, this technology uses Google's speech recognition API.

[0974] The converted text data is then further summarized. For summarization, we use the Hugging Face transformers library. As a concrete example, we use the T5 model to concisely summarize the text content.

[0975] The next step is to use a generative AI model to assess the safety of the summarized text. This assessment is performed using a custom SafetyEvaluator model. The generative AI model is fed a specific prompt, such as "Please assess whether this call is fraud-related. Provide the following summarized text: 'Confirm unusual account activity, request for personal information.'"

[0976] After the evaluation result is obtained, the call connection is controlled based on the result. If the evaluation result is determined to be "safe," the call is successfully connected to the user's device. Conversely, if the evaluation result is "unsafe," the call is blocked and automatically reported to the fraud department. The report contains the necessary information to enable the specialized department to take action quickly.

[0977] In this way, the present invention makes it possible to accurately and quickly evaluate the security of calls and detect and deal with fraudulent calls in real time, thereby protecting users from fraudulent activities and significantly reducing the damage caused by telephone fraud and other crimes.

[0978] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0979] Step 1:

[0980] When a user receives a call, the device's microphone is used to record the call audio. The call audio data is transmitted from the microphone to a server in real time and stored in a digital format. The input is raw audio data, and the output is digital audio data.

[0981] Step 2:

[0982] The server converts the recorded voice data into text data using speech recognition technology. This process uses Google's speech recognition API. The input is digital voice data and the output is text data. The server analyzes the voice data and converts each voice segment into the corresponding text.

[0983] Step 3:

[0984] The server summarizes the converted text data. This summarization utilizes Hugging Face's transformers library, specifically the T5 model. The input is the converted text data, and the output is the summary text. The server analyzes the text data, extracts important information, and summarizes it concisely.

[0985] Step 4:

[0986] The server analyzes the summarized text data and evaluates safety using a generative AI model. Using a custom SafetyEvaluator model, the input is the summarized text and the output is the safety evaluation result. For example, this process inputs a prompt such as, "Please evaluate whether this call is fraud-related. Provide the following summarized text: 'Confirm abnormal account transactions, request for personal information.'"

[0987] Step 5:

[0988] The server controls the call connection based on the evaluation results. If the evaluation result is judged to be "safe," the call is connected to the user's device. The input is the safety evaluation result, and the output is whether the call can be connected. Conversely, if the call is judged to be "unsafe," the call is blocked and an automatic report is sent to the fraud prevention department. The report includes a summary of the specific call content and details of the evaluation result.

[0989] These processing steps enable accurate and rapid assessment of call security and real-time detection and response to fraudulent calls.

[0990] 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.

[0991] The present invention provides a system that records the contents of phone calls, converts the voice data into text, then summarizes it, and evaluates the safety of the call using a generative AI and emotion engine. Below, we will explain the program processing and specific examples of this system.

[0992] System configuration and processing

[0993] 1. Call logging

[0994] When the user answers the phone, the server uses a call recording method to capture the user's voice through a microphone, and this voice data is stored on the server as is.

[0995] 2. Transcription of audio data

[0996] The server uses the acquired voice data to activate a means for converting voice to text, which is performed using voice recognition technology to obtain the contents of the call as text data.

[0997] 3. Summary of Contents

[0998] The server then uses a means for summarizing the converted text data to concisely summarize the contents of the text, allowing the user to obtain the necessary information in a condensed form.

[0999] 4. AI-powered filtering

[1000] The server then invokes the generative AI to analyze the summarized text. The AI ​​uses natural language processing techniques to analyze the text and assess whether the call is related to fraud or other misconduct, using historical data and pattern recognition.

[1001] 5. Emotion Recognition by Emotion Engine

[1002] The server uses the user's voice data to activate the emotion engine, which uses voice analysis technology to assess the user's emotions and stress level, thereby detecting emotions such as confusion or nervousness.

[1003] 6. Integration of evaluation results and call connection control

[1004] The server combines the evaluation results of the generative AI and the emotion engine to perform a comprehensive safety assessment of the call content. If the assessment result is deemed "safe," the server connects the call to the user's device.

[1005] If the evaluation result is judged to be "risky," the server will not connect the call to the user's device and will report it to the fraud department.

[1006] Specific examples

[1007] For example, consider the case where an elderly person receives a phone call from someone claiming to be from a bank. The call is first recorded by a server. The recorded voice data is then converted into text. The converted text might say something like, "We've recently observed unusual transactions from your account. Please provide your personal information to verify."

[1008] The server summarizes this text into a short summary such as "Abnormal account transaction confirmation, request for personal information." The summarized text is then evaluated by a generative AI, which in this case determines it to be "risky."

[1009] At the same time, the emotion engine analyzes the voice data and determines whether the user is confused or nervous. This information is also added to the overall evaluation. As a result, the server will not connect the call to the user's device and will report it to the fraud department.

[1010] This system can significantly reduce the risk of elderly people becoming victims of telephone fraud. In addition, by taking into account the user's emotions, it can provide a more accurate evaluation.

[1011] The processing flow will be explained below.

[1012] Step 1:

[1013] The user answers the call. The server activates the call recording function and captures the user's voice in real time through the microphone. This voice data is stored as is on the server.

[1014] Step 2:

[1015] The server uses the recorded voice data to activate a voice-to-text conversion means. Specifically, it uses voice recognition technology to convert the voice data into text data. The result of this conversion is a text representation of the contents of the call.

[1016] Step 3:

[1017] The server uses a means to summarize the converted text data, turning long conversations into concise summaries. The summarization process extracts important information and summarizes it in an easy-to-understand format.

[1018] Step 4:

[1019] The server then invokes the generative AI to analyze the summarized text. The AI ​​uses natural language processing techniques to analyze the text and assess whether the call is related to fraud or other misconduct, using historical data and pattern recognition.

[1020] Step 5:

[1021] The server receives the evaluation results from the AI ​​generator, along with the user's voice data, which then activates the emotion engine. The emotion engine uses voice analysis technology to evaluate the user's emotions and stress level, detecting whether the user is confused, nervous, or relieved.

[1022] Step 6:

[1023] The server combines the results of the generative AI and the emotion engine to perform a comprehensive safety assessment of the call content, specifically determining whether the likelihood of fraud is high or low, and whether the user is nervous.

[1024] Step 7:

[1025] If the evaluation result is deemed "safe," the server connects the call to the user's device, allowing the user to continue the call with peace of mind.

[1026] Step 8:

[1027] If the evaluation results in a "risk" rating, the server will not connect the call to the user's device and will instead report it to the fraud department. This report will include the specific content of the call and information about the user's emotional state, allowing for a prompt response.

[1028] Specific examples

[1029] For example, consider the case where an elderly person receives a phone call from someone claiming to be from their bank. The call is first recorded by a server. The recorded voice data is then converted into text. The converted text might say something like, "We've recently observed unusual transactions from your account. Please provide your personal information to verify."

[1030] The server summarizes this text into a short summary such as "Abnormal account transaction confirmation, request for personal information." The summarized text is then evaluated by a generative AI, which in this case determines it to be "risky."

[1031] At the same time, the emotion engine analyzes the voice data and evaluates whether the user is confused or nervous. This information is also taken into account in the overall evaluation. As a result, the server reports the call to the fraud department without connecting it to the user's device. This system can significantly reduce the risk of elderly people becoming victims of telephone fraud. Furthermore, taking the user's emotions into account allows for more accurate evaluation.

[1032] Example 2

[1033] 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."

[1034] In recent years, telephone fraud targeting the elderly has been increasing, and an effective system to prevent such fraud is required. To prevent such fraud, it is necessary to analyze the contents of calls quickly and accurately and to identify fraudulent or dangerous calls in real time. Conventional systems do not take into account the user's emotions or stress state, making it difficult to improve accuracy.

[1035] 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.

[1036] In this invention, the server includes means for recording voice, means for converting the recorded voice into text, means for summarizing the converted text, means for analyzing the summarized text using a generative AI model to assess the risk of fraud or misconduct, means for analyzing the recorded voice using an emotion analysis engine to assess the emotional state, means for integrating the evaluation results of the generative AI model and the emotion analysis engine to comprehensively assess the safety of communication, and means for controlling call connections based on the evaluation results. This makes it possible to assess the risk of telephone fraud with high accuracy and comprehensively determine the safety of calls while taking the user's emotional state into consideration.

[1037] "Audio recording means" refers to devices or techniques used to capture the contents of telephone conversations.

[1038] The "means for converting recorded voice into text" refers to a voice recognition technology for converting voice data into text data.

[1039] The "means for summarizing converted characters" is a technique for concisely summarizing text data that has been recognized by speech recognition.

[1040] "Means for analyzing using a generative AI model to assess the risk of fraud or misconduct" refers to a technology that uses AI technology to analyze text data and assess the possibility of fraud or misconduct.

[1041] The "means for analyzing using an emotion analysis engine and assessing the emotional state" is a technology that analyzes voice data to assess the user's emotions and stress state.

[1042] "Means for integrating evaluation results and comprehensively assessing the safety of communication" refers to a technology that integrates the evaluation results obtained from a generative AI model and an emotion analysis engine to comprehensively judge the safety of a call.

[1043] The "means for controlling call connection based on the evaluation result" is a technology for controlling call connection based on the safety evaluation result of the call content.

[1044] This invention relates to a system that records phone conversations, converts the audio data into text, then summarizes it and evaluates the safety of the call using generative AI and an emotion engine.

[1045] First, when a user answers a phone call, the contents of the call are recorded by the device. The recorded voice data is immediately sent to the server and stored on the server side. Specifically, the user's voice is picked up through the device's microphone, and the voice data is transferred to the server using a network protocol (e.g., UDP or TCP). The server then stores this data in cloud storage (e.g., Amazon S3).

[1046] The server then converts the recorded audio data into text data using speech recognition technology, specifically a speech recognition engine (e.g., Google Cloud Speech-to-Text API). The server sends the audio file to the API and stores the returned text data in a database (e.g., MySQL).

[1047] The server then uses a summarization engine (e.g., OpenAI GPT-3) to summarize the converted text data. The server sends the text data along with the prompt to the summarization engine and receives the summarized text data. The summary results are also stored in a database.

[1048] The server then uses a generative AI model to perform a risk assessment on the summarized text. Specifically, the summarized text is sent to a generative AI (e.g., OpenAI GPT-3) again to assess the likelihood of fraud or misconduct. The results of this assessment are also stored on the server.

[1049] Furthermore, the server launches an emotion analysis engine (e.g., IBM Watson Tone Analyzer) to recognize and evaluate the user's emotions from the recorded voice data. The server sends the voice data to the emotion analysis engine and receives the analysis results. These evaluation results are also stored in the database.

[1050] Finally, the server combines the results of the generative AI model and the sentiment analysis engine to evaluate the overall safety of the call. If the evaluation result is "safe," the server connects the call to the user's device. On the other hand, if the evaluation result is "risky," the server notifies the fraud department and blocks the call.

[1051] Specific examples

[1052] For example, if an elderly person receives a call from someone claiming to be from a bank, the system operates as follows:

[1053] 1. When a user receives a call, the device captures the voice data and sends it to the server.

[1054] 2. The server sends the audio data to the Google Cloud Speech-to-Text API and converts it into text data.

[1055] 3. The converted text data is summarized using OpenAI GPT-3. For example, a message such as "We have recently observed unusual transactions from your account. Please provide your personal information to confirm" is summarized as "Confirming unusual transactions on your account, requesting the provision of personal information."

[1056] 4. The server analyzes this summary text using OpenAI GPT-3 and rates it as "dangerous."

[1057] 5. At the same time, IBM Watson Tone Analyzer is used to analyze the user's emotions and assess whether the user is confused or nervous.

[1058] 6. The server combines these evaluation results and, if it determines the call is "risky," it reports it to the fraud department and blocks the call.

[1059] Prompt Sentence Examples

[1060] "If you receive a call from someone claiming to be a bank employee and they ask for personal information, transcribe the call and then summarize it to assess its security."

[1061] The present invention allows for safe and reliable call management by effectively assessing the risk of telephone fraud and taking into account the emotional state of the user.

[1062] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1063] Step 1:

[1064] When a user answers a call and the device starts a call, the server captures and receives the voice data in real time. Specifically, the device's microphone captures the user's voice and sends the voice data to the server using UDP or TCP protocols. The server then stores the received voice data in storage (e.g., Amazon S3).

[1065] Input: User's voice data

[1066] Output: Audio file saved to storage

[1067] Step 2:

[1068] The server retrieves the voice data from the storage and converts it into text data using voice recognition technology. It then activates a voice recognition engine (e.g., Google Cloud Speech-to-Text API) and sends the voice data to a cloud service. The converted text data is then stored in a database (e.g., MySQL).

[1069] Input: Audio file

[1070] Output: Character data stored in the database

[1071] Step 3:

[1072] The server retrieves text data from the database and condenses it using summarization technology. It then launches a summarization engine (e.g., OpenAI GPT-3) and sends the text data along with a prompt to the summarization engine. The summarized text data is then stored in the database.

[1073] Input: Character data

[1074] Output: Summary text stored in the database

[1075] Step 4:

[1076] The server analyzes the summarized text using a generative AI model to assess the risk of fraud and misconduct. The summarized text, along with a prompt, is sent to the generative AI (e.g., OpenAI GPT-3) to obtain a risk assessment result, which is also stored in a database.

[1077] Input: Summary text

[1078] Output: Risk assessment results stored in a database

[1079] Step 5:

[1080] The server passes the voice data to an emotion analysis engine to analyze the user's emotion. It then starts the emotion analysis engine (e.g., IBM Watson Tone Analyzer) and sends the voice data. The analysis results are stored in a database.

[1081] Input: Audio file

[1082] Output: Sentiment analysis results stored in a database

[1083] Step 6:

[1084] The server combines the evaluation results of the generative AI model and the sentiment analysis engine to perform a comprehensive safety assessment of the call content. The evaluation results are input into the integrated algorithm to perform a final safety assessment. If the evaluation result is judged to be "safe," the server instructs the call to be connected to the user's device. If the evaluation result is judged to be "risky," the server reports the call to the fraud department and blocks the call.

[1085] Input: Risk assessment results, sentiment analysis results

[1086] Output: Connect or block the call and report it to the fraud department

[1087] (Application example 2)

[1088] 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."

[1089] In modern society, telephone fraud is on the rise. The elderly and general public are especially at risk of falling prey to fraudulent calls, and traditional manual countermeasures have their limitations. There is also a need for systems that can automatically analyze the content of phone calls, evaluate their safety, and provide feedback and alerts to users.

[1090] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for recording voice, means for converting the recorded voice into text, means for summarizing the converted text, means for analyzing the summarized text using a generation AI and an emotion engine to evaluate the safety of communication and the user's emotions, means for controlling call connection based on the evaluation results, and means for feeding back the evaluation results to the user. This makes it possible to evaluate the safety of calls in real time, provide quick feedback to the user, and reduce the risk of fraud.

[1091] An "audio recording means" is a device or system that records telephone conversations in real time and stores them as digital audio data for subsequent processing.

[1092] A "means for converting recorded voice into text" is a device or system that uses voice recognition technology to convert recorded voice data into text data.

[1093] The "means for summarizing converted characters" is a device or system that analyzes text data, extracts important information, and summarizes it concisely.

[1094] "Means for analyzing using generative AI and an emotion engine to assess the safety of communication and the user's emotions" refers to a device or system that uses a generative AI model and an emotion analysis engine to assess emotions from the summarized text content and voice, and determine the safety of the call and the user's emotional state.

[1095] "Means for controlling call connections based on evaluation results" refers to a device or system that connects a call if it is judged to be safe based on the evaluation results of the generative AI and emotion engine, and blocks the connection if it is judged to be dangerous.

[1096] "Means for providing feedback of evaluation results to the user" refers to a device or system that notifies the user of the evaluation results by the generative AI and emotion engine, and provides information on the safety of the call and the connection status.

[1097] A "means for reporting to the fraud department" is a device or system that, if a call is deemed dangerous, notifies the fraud department of that information and prompts them to take appropriate action.

[1098] To implement this invention, the following system configuration and programs are required: The entire system is built around three elements: a server, a terminal, and a user.

[1099] System Configuration

[1100] 1. Server

[1101] As a means of recording audio, the server captures audio data during a call in real time and stores it as digital data.

[1102] To convert the recorded voice into text, the server uses voice recognition technology to convert the acquired voice data into text data. Specifically, it uses the speech_recognition library.

[1103] To summarize the converted text, the server uses natural language processing techniques to extract and summarize important information from the text data, for example, by using spacy to analyze the content of the text.

[1104] The summarized text is analyzed by the generative AI and emotion engine, and the generative AI model and emotion engine are activated as a means of evaluating the safety of communication and the user's emotions. The generative AI model uses a model from the transformers library, and the emotion engine uses the emotion_recognition library.

[1105] As a means of controlling call connections based on the evaluation results, the server summarizes the evaluation results from the AI ​​and emotion engine, and connects the call if it is judged to be safe, and stops the connection if it is judged to be dangerous.

[1106] As a means of providing feedback on the evaluation results to the user, the evaluation results are sent to the terminal and notified to the user.

[1107] 2. Terminal

[1108] The device receives the evaluation results sent from the server and provides visual or auditory feedback to the user, for example by displaying a warning message using the smartphone's notification function.

[1109] 3. Users

[1110] During a telephone call, the user receives feedback from the terminal and decides whether to continue or terminate the call as needed.

[1111] Specific examples

[1112] For example, if a user receives a call from someone claiming to be from a "bank," the call is first recorded by the server. The recorded voice data is converted into text data and recognized as, "We have recently observed unusual activity on your account. Please provide your personal information for verification." The server summarizes this text and summarizes it as, "Abnormal activity on account, request for personal information." This summarized text is then analyzed using a generative AI model and an emotion engine, and is rated as "dangerous." This evaluation result is sent to the device and displayed as a warning to the user. As a result, the user recognizes that the call may be fraudulent and can take appropriate action.

[1113] Prompt Sentence Examples

[1114] The following prompts are fed into the generative AI model for evaluation:

[1115] "Please summarize the following call and assess its likelihood of fraud or misconduct.\nCall content: {Call content}"

[1116] Thus, the present invention provides an effective system for reducing the risk of telephone fraud and keeping users safe.

[1117] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1118] Step 1:

[1119] When the server detects that the user has started a phone call, it starts recording the audio in real time by capturing the audio data using a microphone and storing the digital data on the server.

[1120] Input: User's call audio

[1121] Output: Digital audio data

[1122] Step 2:

[1123] The server converts the recorded voice data into text data using speech recognition technology. Specifically, it uses the speech_recognition library to convert voice to text.

[1124] Input: Digital audio data

[1125] Output: Text data

[1126] Step 3:

[1127] The server summarizes the text data using natural language processing techniques, such as the spacy library, to extract and summarize key information concisely.

[1128] Input: Text data

[1129] Output: Summarized text data

[1130] Step 4:

[1131] The server runs a generative AI model and an emotion engine to analyze the content of the summarized text and assess the safety of the call and the user's emotions. The generative AI model uses the transformers library, and the emotion engine uses the emotion_recognition library.

[1132] Input: Summarized text data

[1133] Output: Safety evaluation results and emotion evaluation results

[1134] Step 5:

[1135] The server controls the call connection based on the evaluation results of the generative AI and emotion engine. If the call is judged to be safe, the connection is maintained, and if it is judged to be dangerous, the connection is cut off.

[1136] Input: Safety evaluation results and emotion evaluation results

[1137] Output: Call connection status (connected / disconnected)

[1138] Step 6:

[1139] The server sends the evaluation results to the device and notifies the user, so the user can immediately know whether the call is safe or dangerous. The device displays this using its notification function.

[1140] Input: Safety evaluation results and emotion evaluation results

[1141] Output: User notification

[1142] Step 7:

[1143] The user receives feedback from the device and decides whether to continue or terminate the call, and if necessary, terminates the call and reports the call to the fraud department.

[1144] Input: User notification

[1145] Output: User behavior (continue / cancel call, report)

[1146] 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.

[1147] 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.

[1148] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1149] 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.

[1150] 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.

[1151] 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.

[1152] 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).

[1153] 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.

[1154] 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."

[1155] 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.

[1156] 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).

[1157] 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.

[1158] 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.

[1159] 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.

[1160] 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.

[1161] 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.

[1162] 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.

[1163] 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.

[1164] 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.

[1165] 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.

[1166] 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.

[1167] The following is further disclosed regarding the above embodiment.

[1168] (Claim 1)

[1169] a means for recording audio;

[1170] a means for converting the recorded audio into text;

[1171] a means for summarizing the converted characters;

[1172] a means for analyzing the summarized text and assessing the security of the communication;

[1173] means for controlling call connection based on the evaluation result;

[1174] A system including:

[1175] (Claim 2)

[1176] 2. The system of claim 1, further comprising means for connecting the call only if the evaluation result indicates safety.

[1177] (Claim 3)

[1178] 10. The system of claim 1, further comprising means for reporting the call to a fraud department if the evaluation determines that the call is unsafe.

[1179] "Example 1"

[1180] (Claim 1)

[1181] a means for recording calls;

[1182] means for converting the recorded audio of the call into text;

[1183] a means for summarizing the converted characters;

[1184] A means of analyzing the summarized text and assessing the security of the call content using a generative AI model;

[1185] means for controlling call connection based on the evaluation result;

[1186] A system including:

[1187] (Claim 2)

[1188] 2. The system of claim 1, further comprising means for connecting the call only if the evaluation result indicates safety.

[1189] (Claim 3)

[1190] 10. The system of claim 1, further comprising means for reporting the call to a supervisory response department if the evaluation results in an unsafe call.

[1191] "Application Example 1"

[1192] (Claim 1)

[1193] a means for recording audio;

[1194] a means for converting the recorded audio into text;

[1195] a means for summarizing the converted characters;

[1196] A means for analyzing the summarized text and using a generative AI model to assess the safety of the communication; and

[1197] a means for controlling call connections based on the evaluation result and reporting to an anti-fraud department if the evaluation is unsafe;

[1198] A system including:

[1199] (Claim 2)

[1200] 2. The system of claim 1, further comprising means for connecting the call only if the evaluation result indicates safety.

[1201] (Claim 3)

[1202] 10. The system of claim 1, further comprising means for reporting the call to a fraud department if the evaluation determines that the call is unsafe.

[1203] "Example 2: Combining Emotion Engines"

[1204] (Claim 1)

[1205] a means for recording audio;

[1206] a means for converting the recorded audio into text;

[1207] a means for summarizing the converted characters;

[1208] A means of analyzing the summarized text using a generative AI model to assess the risk of fraud and misconduct;

[1209] means for analyzing the recorded voice using an emotion analysis engine to assess the emotional state;

[1210] A means to comprehensively evaluate the safety of communication by integrating the evaluation results of the generative AI model and the emotion analysis engine, and

[1211] means for controlling call connection based on the evaluation result;

[1212] A system including:

[1213] (Claim 2)

[1214] 2. The system of claim 1, further comprising means for connecting the call only if the evaluation result indicates safety.

[1215] (Claim 3)

[1216] 10. The system of claim 1, further comprising means for reporting the call to a fraud department if the evaluation determines that the call is unsafe.

[1217] "Application example 2 when combining emotion engines"

[1218] (Claim 1)

[1219] a means for recording audio;

[1220] a means for converting the recorded audio into text;

[1221] a means for summarizing the converted characters;

[1222] A means for analyzing the summarized text using a generative AI and an emotion engine to evaluate the safety of communication and the user's emotions;

[1223] means for controlling call connection based on the evaluation result;

[1224] a means for providing feedback of the evaluation results to the user;

[1225] A system including:

[1226] (Claim 2)

[1227] 2. The system according to claim 1, further comprising means for connecting the call only if the evaluation result indicates safety, and notifying the user.

[1228] (Claim 3)

[1229] 10. The system of claim 1, further comprising means for alerting the user and reporting the call to a fraud department if the evaluation determines that the call is unsafe. [Explanation of symbols]

[1230] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. a means for recording audio; a means for converting the recorded audio into text; a means for summarizing the converted characters; a means for analyzing the summarized text and assessing the security of the communication; means for controlling call connection based on the evaluation result; A system including:

2. 2. The system of claim 1, further comprising means for connecting the call only if the evaluation result indicates that the call is safe.

3. 10. The system of claim 1, further comprising means for reporting the call to a fraud department if the evaluation determines that the call is unsafe.

Images