system

Abstract

This system receives call audio data, converts it to text data, analyzes it, and transmits the generated response audio data based on the analysis results. [Solution] A means for receiving audio data and converting it into text data in real time, A means for analyzing text data and converting the response generated based on the analysis results into audio data, A system including means for transmitting the generated voice data of the response. Furthermore, it includes means to allow the user to intervene in a call if specific keywords are detected in the text.

Description

Technical Field

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

Background Art

[0002] Patent Document 1 discloses a method for controlling a persona chatbot, which is performed by at least one processor, including steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to an explanation of a character of the chatbot, 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

Summary of the Invention

Problems to be Solved by the Invention

[0004] In a modern business environment, operations such as sales calls and claim handling by phone require a large amount of manpower, consuming labor and time, and there is a problem that one cannot concentrate on important operations. In the current system, since the automation of phone response is not sufficiently carried out, it is difficult for users to efficiently perform operations. Therefore, there is a need for a new system to improve the efficiency of phone response and reduce the workload of users.

Means for Solving the Problems

[0005] The present invention provides a system that includes means for receiving voice data and converting it into text data in real time, means for analyzing the text data and converting the response generated based on the analysis results into voice data, and means for transmitting the voice data of the generated response. This system automates telephone answering and reduces the effort required from the user. Furthermore, by including means for detecting specific keywords from the text data and notifying the user, it enables user intervention at important points and realizes a seamless call flow. In addition, by including means for allowing the user to intervene in the call when specific keywords are detected, flexible responses that combine human judgment as needed become possible.

[0006] "Voice data" refers to voice information in analog or digital format acquired through phone calls or voice input.

[0007] "Text data" refers to string-formatted information generated by analyzing audio data.

[0008] "Real-time" refers to a processing mode in which input information is processed immediately and results are output without delay.

[0009] "Analysis" is an information processing method used to process received data and understand its content and intent.

[0010] "Generated response" refers to a reply message that is automatically generated based on the analysis results.

[0011] "Converting to audio data" refers to the process of converting text data into an audio format.

[0012] "Transmission" refers to the process of transferring generated audio data to the intended recipient or device.

[0013] A "specific keyword" is a particular word or phrase that the system deems important.

[0014] A "notification" is a means of informing users of information or warnings.

[0015] "Intervention" refers to the act of a user stopping the automatic response and participating in the call to handle it themselves.

[0016] "Pre-set conditions" are pre-defined situations or criteria necessary for the system to perform specific operations.

Brief Description of Drawings

[0017] [Figure 1] It is a conceptual diagram showing an example of the configuration of a data processing system according to the first embodiment. [Figure 2] It is a conceptual diagram showing an example of the main functions of a data processing device and a smart device according to the first embodiment. [Figure 3] It is a conceptual diagram showing an example of the configuration of a data processing system according to the second embodiment. [Figure 4] It is a conceptual diagram showing an example of the main functions of a data processing device and smart glasses according to the second embodiment. [Figure 5] It is a conceptual diagram showing an example of the configuration of a data processing system according to the third embodiment. [Figure 6] It is a conceptual diagram showing an example of the main functions of a data processing device and a headset-type terminal according to the third embodiment. [Figure 7] It is a conceptual diagram showing an example of the configuration of a data processing system according to the fourth embodiment. [Figure 8] It is a conceptual diagram showing an example of the main functions of a data processing device and a robot according to the fourth embodiment. [Figure 9] It shows an emotion map to which multiple emotions are mapped. [Figure 10] It shows an emotion map to which multiple emotions are mapped. [Figure 11] It is a sequence diagram showing the processing flow of the data processing system in Example 1. [Figure 12]It is a sequence diagram showing the processing flow of the data processing system in Application Example 1. [Figure 13] It is a sequence diagram showing the processing flow of the data processing system in Example 2 when the emotion engine is combined. [Figure 14] It is a sequence diagram showing the processing flow of the data processing system in Application Example 2 when the emotion engine is combined.

Mode for Carrying Out the Invention

[0018] Hereinafter, an example of an embodiment of the system according to the technology of the present disclosure will be described with reference to the accompanying drawings.

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

[0020] In the following embodiments, the numbered processor (hereinafter simply referred to as "processor") may be a single arithmetic unit or a combination of multiple arithmetic units. Also, the processor may be a single type of arithmetic unit or a combination of multiple types of arithmetic units. Examples of arithmetic units include CPU (Central Processing Unit), GPU (Graphics Processing Unit), GPGPU (General-Purpose computing on Graphics Processing Units), APU (Accelerated Processing Unit), etc.

[0021] In the following embodiments, the numbered RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a work memory by the processor.

[0022] In the following embodiments, the signed storage is one or more non-volatile storage devices that store various programs and various parameters. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), or magnetic tapes.

[0023] In the following embodiments, the signed communication interface (I / F) is an interface that includes a communication processor and an antenna, etc. The communication interface manages communication between multiple computers. Examples of communication standards applicable to the communication interface include wireless communication standards such as 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), or Bluetooth (registered trademark).

[0024] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." That is, "A and / or B" means that it may be A alone, or B alone, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" applies when expressing three or more things linked by "and / or."

[0025] [First Embodiment]

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

[0027] As shown in Figure 1, the data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0028] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0029] The smart device 14 comprises a computer 36, a reception device 38, an output device 40, a camera 42, and a communication interface 44. The computer 36 comprises a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The reception device 38, output device 40, and camera 42 are also connected to the bus 52.

[0030] The reception device 38 is equipped with a touch panel 38A and a microphone 38B, etc., and receives user input. The touch panel 38A receives user input by detecting contact with an object (e.g., a pen or finger). The microphone 38B receives user input by detecting the user's voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0031] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form perceptible to the user 20 (e.g., audio and / or text). The display 40A displays visible information such as text and images according to instructions from the processor 46. The speaker 40B outputs audio according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0032] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various types of information between processor 46 and processor 28 via network 54.

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

[0034] As shown in Figure 2, in the data processing device 12, a specific processing 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" related to the technology of this 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 according to the specific processing program 56 executed on the RAM 30.

[0035] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0036] In the smart device 14, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The reception output program 60 is used in conjunction with a 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0037] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0038] The system according to the present invention receives voice data, converts it to text data in real time, analyzes it, generates an appropriate response, and transmits it as voice data. It also detects specific keywords and notifies the user, enabling seamless switching so that the user can intervene as needed.

[0039] System-wide configuration

[0040] The system consists of the following elements:

[0041] 1. Terminal

[0042] 2. Server

[0043] 3. User

[0044] Receiving and converting audio data

[0045] The device detects an incoming call and, as soon as the call begins, sends the audio data to the server in real time.

[0046] The server passes the received audio data to the speech recognition engine, which converts it into text data in real time.

[0047] Text data analysis and response generation

[0048] The server uses a natural language processing (NLP) engine to analyze text data and understand its content.

[0049] Based on the analysis results, the server automatically generates an appropriate response using a generation AI. This response is then generated again in text format.

[0050] The server passes the text-based response to the speech synthesis engine, which converts it into speech data.

[0051] Sending a response and notification

[0052] The server sends the generated voice data to the terminal, which then relays it to the person on the other end of the line.

[0053] The server continuously monitors the call content and notifies the user via the terminal if it detects specific keywords.

[0054] Seamless user intervention

[0055] Users can check notifications and, if important call content is detected, intervene in the call themselves.

[0056] The device accepts user intervention, allowing the user to speak directly with the person they are calling.

[0057] Specific example

[0058] 1. Handling sales calls

[0059] A salesman makes a phone call and says, "Excuse me for bothering you while you're busy, this is a salesperson from AA Insurance."

[0060] The device transmits voice data to the server in real time.

[0061] The server converts the voice data into text and recognizes it as "This is a salesperson from AA Insurance."

[0062] The server uses a generation AI to generate the response, "I'm not interested, thank you," converts it into audio data, and sends it to the terminal.

[0063] The device responds to the salesman by playing this audio data.

[0064] 2. Important notifications and user intervention

[0065] The salesman continued, "Would you mind if I sent you some materials?"

[0066] The server detects specific keywords such as "send documents" and notifies the user via the terminal.

[0067] The user checks the notification and taps the screen to intervene in the call, directly responding, "I don't need to send any documents."

[0068] Thus, the present invention is a system that automates telephone answering while allowing user intervention in important situations, thereby achieving efficient and flexible responses.

[0069] Next, we can describe the details of the processing steps.

[0070] The following describes the processing flow.

[0071] Step 1:

[0072] The device detects an incoming call and collects the call audio as soon as the call begins. The collected audio data is transmitted to the server in real time.

[0073] Step 2:

[0074] The server passes the received audio data to the speech recognition engine, which converts the audio data into text data. This conversion is performed in real time and processed sequentially.

[0075] Step 3:

[0076] The server passes the text data to a natural language processing (NLP) engine for content analysis. The analysis identifies the purpose and topic of the call.

[0077] Step 4:

[0078] The server then uses the generation AI to create a response based on the analysis results. This response is generated in text format and designed to be an appropriate reply to the conversation.

[0079] Step 5:

[0080] The server passes the generated text response to the speech synthesis engine, which converts it into speech data. The converted speech data is then generated again in real time.

[0081] Step 6:

[0082] The server sends the audio data to the terminal. The terminal decodes the received audio data and plays it back to the caller at the appropriate time.

[0083] Step 7:

[0084] The server continues to monitor the call content to check for the inclusion of specific keywords or phrases. For example, it may detect specific keywords such as "important matter" or "sending documents."

[0085] Step 8:

[0086] When a specific keyword is detected, the server sends a notification to the terminal, informing the user that intervention is necessary.

[0087] Step 9:

[0088] The device displays a notification to the user, alerting them with audio and visuals. It then prompts the user to choose whether or not to take action.

[0089] Step 10:

[0090] The user checks the notification and, if important call content is detected, intervenes in the call at their own discretion. The device accepts the user's intervention and allows the user to speak directly with the other party.

[0091] Step 11:

[0092] Once the call ends, the device sends the audio recording and text data to the server. The server stores this data in a database for later reference.

[0093] In this way, the system seamlessly integrates real-time voice responses with user intervention, enabling effective telephone handling.

[0094] (Example 1)

[0095] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0096] Conventional voice response systems were unable to adequately generate responses or detect keywords in real time, making it difficult to respond immediately to important call content. Furthermore, the inability to seamlessly integrate automated responses with user intervention made flexible and efficient call handling challenging.

[0097] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0098] In this invention, the server includes means for receiving voice data and converting it into text data in real time; means for analyzing the text data and generating an appropriate response using a generative AI model; means for converting the generated text data of the response into voice data; means for transmitting the generated voice data of the response; means for detecting specific keywords and notifying the user; and means for allowing the user to acknowledge the notification and intervene in the call. This enables real-time text conversion and response generation, keyword detection and user notification, and seamless user intervention.

[0099] "Audio data" refers to audio signals obtained through phone calls, recordings, etc., represented in digital format.

[0100] "Real-time" refers to data processing and responses occurring immediately, with results reflected without delay.

[0101] "Text data" refers to audio data converted into written information, expressed in a format that is understandable to humans.

[0102] A "generative AI model" refers to a technology that uses artificial intelligence algorithms to generate appropriate responses or data based on given input.

[0103] A "prompt" is a series of text inputs to a generative AI model, which the AI ​​uses to generate a response.

[0104] "Analysis" is a means of processing text data to understand its content and structure, and a process for deriving appropriate responses or actions.

[0105] A "keyword" is an important word or phrase contained within text data, and its detection triggers a specific action or response.

[0106] A "user" is a person who uses the system and has the authority to intervene in the content of a call.

[0107] A "notification" is an informational alert sent from the system to the user, intended to draw the user's attention when a specific keyword is detected.

[0108] "Intervening in a call" refers to the action of a user switching from an automated system response to a manual response and starting a direct conversation with the other party.

[0109] A "speech recognition engine" is software or hardware that analyzes speech data and converts it into text data.

[0110] A "speech synthesis engine" is software or hardware that analyzes text data and converts it into speech data.

[0111] "Communication methods" refer to technologies used to send and receive voice and text data between a server and a terminal, and include the Internet Protocol and other communication protocols.

[0112] A specific embodiment of the system according to the present invention is shown below. This system receives voice data, converts it into text data in real time, performs subsequent analysis and response generation, and enables notifications and call interventions to the user.

[0113] Receiving and converting audio data

[0114] The terminal detects an incoming call and captures audio data as soon as the call begins. The call audio is transmitted to the server in real time. WebSocket and HTTP streaming technologies are used as secure communication methods for transmitting the audio data.

[0115] The server converts the received audio data into text data using the Google® Cloud Speech-to-Text API or Amazon Transcribe. The converted text data is extracted as a string that accurately represents the spoken content.

[0116] Text data analysis and response generation

[0117] The server uses Google Cloud Natural Language API and Microsoft® Azure® Text Analytics to analyze text data and understand its content. The information extracted through analysis includes grammatical structure, meaning, and important keywords.

[0118] Based on the analysis results, the server automatically generates an appropriate response using a generative AI model (e.g., GPT-3®). The generative AI model used here has advanced natural language generation capabilities and generates the optimal response from the given prompt sentence.

[0119] The generated response is then converted back into text format and then into speech data using the Google Text-to-Speech API or Amazon Polly.

[0120] Sending a response and notification

[0121] The server sends the generated audio data to the terminal, which then relays that audio data to the person on the other end of the call. This enables real-time responses.

[0122] A server that continuously monitors call content will notify the user via the terminal if it detects a specific keyword, such as the phrase "send documents." The notification system can display the notification on the terminal screen or alert the user with vibration or sound.

[0123] Seamless user intervention

[0124] Upon receiving a notification, the user can review the notification displayed on their device screen and decide whether or not to intervene in the call. If the user chooses to intervene, tapping the notification on their device will switch the system from automatic response to manual response, allowing the user to speak directly with the caller.

[0125] Specific example

[0126] 1. Handling sales calls

[0127] A salesman makes a phone call and says, "Excuse me for bothering you while you're busy, this is a salesperson from AA Insurance."

[0128] The device transmits voice data to the server in real time.

[0129] The server converts the voice data into text and recognizes it as "This is a salesperson from AA Insurance."

[0130] The server uses a generation AI to generate the response, "I'm not interested, thank you," converts it into audio data, and sends it to the terminal.

[0131] The device responds to the salesman by playing this audio data.

[0132] 2. Important notifications and user intervention

[0133] The salesman continued, "Would you mind if I sent you some materials?"

[0134] The server detects specific keywords such as "send documents" and notifies the user via the terminal.

[0135] The user checks the notification, taps the screen to intervene in the call, and directly responds, "I don't need to send any documents."

[0136] Example of a prompt

[0137] "If a salesperson is soliciting insurance and includes the phrase 'send me some materials,' the user should be notified. The response should generate 'I do not need any materials.'"

[0138] This system enables efficient and flexible call handling by transcribing audio in real time, generating appropriate responses, and allowing user intervention at key points.

[0139] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0140] Step 1:

[0141] The device detects an incoming call and captures audio data as soon as the call begins. The audio data is sent to the server in real time. Specifically, it receives the phone's audio in real time as input and sends that audio data to the server via the network as output. In practice, the smartphone outputs a log message saying "Audio capture started" and immediately begins sending the audio data.

[0142] Step 2:

[0143] The server passes the received audio data to a speech recognition engine (e.g., Google Cloud Speech-to-Text API) and converts it into text data in real time. It takes audio data as input and generates text data as output. Specifically, the server outputs a log message saying "Audio data converted to text."

[0144] Step 3:

[0145] The server uses a natural language processing (NLP) engine (e.g., Google Cloud Natural Language API) to analyze text data. It receives text data as input and generates analysis results as output. Analysis includes entity extraction and sentiment analysis. Specifically, the server logs "Text analysis complete: Intent is sales."

[0146] Step 4:

[0147] The server generates an appropriate response based on the analysis results using a generative AI model (e.g., GPT-3). It receives the analysis results and a prompt as input and generates the appropriate response text as output. Specifically, the server logs "Response generation complete." An example of a prompt is: "If a salesman is soliciting insurance and includes the phrase 'send materials,' the user should be notified. The response should generate 'No materials are needed.'"

[0148] Step 5:

[0149] The server passes the generated response text to a text-to-speech engine (e.g., Google Text-to-Speech API) to convert it into speech data. It takes response text as input and generates speech data as output. Specifically, the server logs "Text-to-speech complete".

[0150] Step 6:

[0151] The server sends the generated audio data to the terminal, and the terminal transmits that audio data to the person it is talking to. It receives audio data as input and plays it back to the person it is talking to as output. Specifically, the terminal log displays "Playing response audio".

[0152] Step 7:

[0153] The server continuously monitors the call content and, if it detects a specific keyword (e.g., "send documents"), it notifies the user via the terminal. It receives real-time text data as input and generates a notification as output. Specifically, the server log records "Keyword detected: Send documents" and sends a notification to the terminal.

[0154] Step 8:

[0155] The user can check notifications and, if important call content is detected, intervene in the call themselves. The device receives notifications as input and intervenes in the call as output. Specifically, the user taps the notification on their device and directly responds with "No documents are needed." The device then logs "User intervened in the call."

[0156] (Application Example 1)

[0157] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0158] Traditional content distribution services have faced challenges in providing personalized recommendations based on user preferences and interests, making it difficult to respond quickly and accurately to the information and entertainment users seek. Furthermore, the lack of systems that use AI-generated models to respond to requests has made it difficult to obtain the information users need in real time.

[0159] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0160] In this invention, the server includes means for receiving voice data and converting it into text data in real time; means for analyzing the text data, generating a response using a generative AI model based on the analysis results, and converting the response into voice data; means for transmitting the generated voice data of the response; means for obtaining a request from the user in voice data and sending a prompt sentence to the generative AI model based on the request; and means for suggesting appropriate content to the user based on the generated prompt. This makes it possible to recommend the most suitable content to the user in real time.

[0161] "Audio data" refers to data obtained by converting audio into a digital format.

[0162] "Real-time" refers to processing data instantly and without delay.

[0163] "Text data" refers to digital data obtained by converting audio data into written text.

[0164] A "generative AI model" refers to an artificial intelligence model that generates appropriate responses based on user input data.

[0165] A "prompt" is a portion of the text data input to a generating AI model, and it functions as an instruction or question for the model.

[0166] "Analysis" refers to the process of understanding text data and extracting its meaning.

[0167] "Content" is a general term for information and entertainment that users consume, such as movies, music, and news.

[0168] "Suggestions" refer to the act of recommending appropriate content based on the user's preferences and requests.

[0169] This invention relates to a system that converts audio data into text data in real time, analyzes the text data to generate a response using a generative AI model, and transmits that response back as audio data. Furthermore, it also includes a function to generate prompt sentences for the generative AI model based on user requests and suggest appropriate content.

[0170] System Configuration

[0171] The system consists of the following elements:

[0172] 1. Terminal

[0173] An input device for receiving audio data (e.g., a microphone)

[0174] An output device for playing audio data (e.g., a speaker)

[0175] 2. Server

[0176] Speech recognition engine

[0177] Natural Language Processing (NLP) Engine

[0178] Generative AI models (e.g., GPT-3)

[0179] Speech synthesis engine

[0180] 3. User

[0181] Person making a request by voice

[0182] Receiving and converting audio data

[0183] The terminal receives the user's voice data and sends it to the server in real time. The server uses a speech recognition engine to convert the voice data into text data.

[0184] Text data analysis and response generation

[0185] The server analyzes the converted text data using a natural language processing (NLP) engine to understand its content. Based on the analysis results, it uses a generative AI model to generate an appropriate response, which is then generated again in text format. The generated text is then converted into speech data by a speech synthesis engine.

[0186] Sending a response and suggesting content

[0187] The generated audio data is sent to the device and transmitted to the user through the device's speaker. Furthermore, a request is obtained from the user via the audio data, and a prompt is sent to the generating AI model based on that request. Based on the generated prompt, appropriate content is suggested to the user.

[0188] Specific example

[0189] For example, if a user requests "Tell me your recommendations for the latest movies," the following prompt message will be generated.

[0190] Examples of prompts to input into a generative AI model:

[0191] User request: Please recommend some of the latest movies.

[0192] Appropriate response:

[0193] Based on this prompt, the server sends information to the AI ​​model that generates the appropriate response text. For example, a response such as "As a recent movie recommendation, we suggest the action film XXX" is generated, converted into audio data, and delivered to the user.

[0194] This allows users to receive real-time responses to their requests and to have content that interests them suggested quickly and accurately.

[0195] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0196] Step 1:

[0197] The device receives the user's voice data.

[0198] Input: User's voice data

[0199] Operation: The microphone captures the user's voice and converts the audio data into a digital format.

[0200] Output: Digital audio data

[0201] Step 2:

[0202] The terminal sends the received digital audio data to the server.

[0203] Input: Digital audio data

[0204] Operation: Transmits audio data to the server in real time via the network interface.

[0205] Output: Audio data sent to the server

[0206] Step 3:

[0207] The server uses a speech recognition engine to convert the audio data into text data.

[0208] Input: Audio data

[0209] Operation: Analyzes audio data using a speech recognition engine (e.g., Google Speech Recognition) and converts it to text.

[0210] Output: Text data

[0211] Step 4:

[0212] The server analyzes the text data using a natural language processing (NLP) engine and understands its meaning.

[0213] Input: Text data

[0214] Operation: Uses a natural language processing (NLP) engine to extract meaning from text data.

[0215] Output: Analysis results (user request details)

[0216] Step 5:

[0217] Based on the analysis results, the server generates a response using an AI model.

[0218] Input: Analysis results

[0219] Operation: A prompt is sent to a generative AI model (e.g., GPT-3), and a response text is generated based on it. An example prompt is "User request: Tell me your recommendations for the latest movies. Appropriate response:"

[0220] Output: Response text data

[0221] Step 6:

[0222] The server converts the generated response text into speech data using a speech synthesis engine.

[0223] Input: Response text data

[0224] Operation: Uses a speech synthesis engine (e.g., pyttsx3) to convert response text into speech data.

[0225] Output: Response voice data

[0226] Step 7:

[0227] The server sends the generated response voice data to the terminal.

[0228] Input: Response voice data

[0229] Operation: Transmits voice data to the terminal in real time via the network interface.

[0230] Output: Response voice data sent to the terminal

[0231] Step 8:

[0232] The device plays back the received voice response data and conveys it to the user.

[0233] Input: Response voice data

[0234] Operation: Plays audio data through the speaker.

[0235] Output: The user receives a voice response.

[0236] Step 9:

[0237] If a new request comes from the user, the same process is repeated.

[0238] Input: New voice request

[0239] Action: Repeat the process from step 1.

[0240] Output: New response to the user

[0241] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0242] The system related to this invention automates telephone answering and allows for seamless user intervention as needed. Furthermore, this system incorporates an emotion engine that recognizes the user's emotions and adjusts its response accordingly.

[0243] System-wide configuration

[0244] The system consists of the following elements:

[0245] 1. Terminal

[0246] 2. Server

[0247] 3. User

[0248] 4. Emotional Engine

[0249] Receiving and converting audio data

[0250] The device detects an incoming call and sends audio data to the server in real time when the call begins. The audio data is the content of the call and is sent to the server sequentially.

[0251] Text data analysis and response generation

[0252] The server receives the audio data and converts it into text data in real time using a speech recognition engine.

[0253] The server passes the text data to the NLP engine, which analyzes the content and intent of the conversation. This clarifies the specific topic and purpose of the call.

[0254] Based on the analysis results, the server uses a generative AI to generate an appropriate response in text format. This process builds an automated response system for calls.

[0255] The server passes the generated text response to the speech synthesis engine, which converts it into speech data in real time.

[0256] Emotion recognition by an emotion engine

[0257] The server passes text and audio data to the emotion engine, which analyzes the user's emotions. For example, emotions can be inferred from the tone of voice and word choice.

[0258] Based on the emotions recognized by the emotion engine, the server adjusts its response. If the user is angry, a more careful response will be generated.

[0259] The server detects specific emotions and keywords and sends a notification to the user based on them. In this case, it is determined that the user needs to intervene.

[0260] Sending a response and notification

[0261] The server sends the generated audio data to the terminal, which then sends it back to the person on the other end of the call. Simultaneously, if important conversation content or emotions are detected, the user is notified.

[0262] The device displays a notification to the user and offers them an option to intervene. If the user chooses to intervene, the call is seamlessly handed over to the user.

[0263] Specific example

[0264] 1. Examples of handling customer complaints

[0265] A customer calls to complain: "The item I ordered hasn't arrived yet!"

[0266] The device sends the audio data to the server.

[0267] The server converts the audio data into text and analyzes it using an NLP engine.

[0268] The server detects keywords such as "order," "not received," and "dissatisfaction," and uses a generation AI to generate a response such as "We apologize, we will check on this immediately."

[0269] The emotion engine detects an angry tone and carefully adjusts its response accordingly.

[0270] The server sends the response to the customer via the terminal.

[0271] The server sends a notification to the user based on sentiment data and keywords. It displays a message saying, "We are handling a complaint and intervention is required."

[0272] The user reviews the notification and selects intervention. The system responds directly to the customer with, "We will investigate immediately, please wait."

[0273] In this way, the system of the present invention incorporates emotion recognition functionality to provide appropriate responses based on the user's emotions, and allows the user to intervene as needed. This enables improved efficiency and quality in telephone communication.

[0274] The following describes the processing flow.

[0275] Step 1:

[0276] The device detects incoming calls and collects call audio as soon as the call begins. The collected audio data is sent to the server in real time.

[0277] Step 2:

[0278] The server passes the received audio data to the speech recognition engine, which converts the audio data into text data. This conversion is performed in real time and processed sequentially.

[0279] Step 3:

[0280] The server further passes the text data to a natural language processing (NLP) engine to analyze its content. The purpose and topic of the call are identified in the analysis results.

[0281] Step 4:

[0282] The server passes the analyzed text data to an emotion engine to analyze the user's emotion. The emotion engine identifies the user's emotion from the tone of voice and word choice.

[0283] Step 5:

[0284] The server integrates the results of the emotion engine and the analysis results of the NLP engine to have the generative AI create an appropriate response. This response is generated in text form and is designed to be an appropriate reply in the conversation.

[0285] Step 6:

[0286] The server passes the generated text response to a text-to-speech engine to convert it into voice data. The generated voice data is generated again in real time. Step 7:

[0288] The server sends the generated voice data to the terminal, and the terminal sends it to the phone call partner. Here, the response content is conveyed to the partner.

[0289] Step 8:

[0290] The server continues to monitor the call content and sends a notification to the user if specific keywords or specific emotions are detected.

[0291] Step 9:

[0292] <0, The terminal displays the notification to the user and conveys a warning to the user audibly or visually. The notification includes a message such as "The customer is showing dissatisfaction. Intervention is required." ​

[0293] Step 10:

[0294] The user reviews the notification and, if important call content or emotions are detected, intervenes in the call at their own discretion. The device accepts the user's intervention.

[0295] Step 11:

[0296] The user chooses to intervene and begins a direct conversation with the person on the other end of the line. This allows for a seamless integration of human judgment into responses, even when emotional responses are required.

[0297] Step 12:

[0298] Once the call ends, the device sends the audio recording and text data to the server. The server stores this data in a database for later reference.

[0299] By combining this with an emotion engine, it becomes possible to automate telephone answering while still providing flexible responses based on the user's emotions.

[0300] (Example 2)

[0301] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0302] Traditional telephone answering systems often relied on manual processes for analyzing call content and generating appropriate responses, resulting in challenges in response quality and efficiency. Furthermore, they struggled to respond in a way that considered user emotions, hindering responses that required quick and appropriate action, particularly in situations like complaint handling. Additionally, delays in detecting specific key keywords or emotions could lead to delayed user intervention and inadequate responses.

[0303] The specific processing by the specific processing unit 290 of the data processing apparatus 12 in Example 2 is realized by the following means.

[0304] In this invention, the server includes means for detecting an incoming call of a telephone, transmitting voice data in real time when the call starts, receiving the voice data, converting it into text data in real time using recognition means, passing the text data to analysis means, creating a response generated based on the analysis result by generation means, passing the text data of the generated response to synthesis means, converting it into voice data, and transmitting the generated voice data. Thereby, real-time analysis of call content and automatic generation of appropriate responses become possible.

[0305] The "means for detecting an incoming call of a telephone" refers to a device or software that identifies a call request received via a telephone line or Internet communication.

[0306] The "means for transmitting voice data in real time" refers to a device or software having a function of immediately processing voice information spoken during a call and transmitting it to other devices or systems through a network.

[0307] The "means for receiving voice data and converting it into text data in real time using recognition means" refers to a device or software having a function of immediately converting received voice information into character data using a voice recognition engine.

[0308] The "means for passing text data to analysis means and creating a response generated based on the analysis result by generation means" refers to a device or software having a function of sending text data to a natural language processing engine and creating an appropriate response sentence based on the analysis result using an AI or the like.

[0309] The "means for passing the text data of the generated response to synthesis means and converting it into voice data" refers to a device or software having a function of sending a response in text form to a voice synthesis engine and converting it into a form that can be reproduced as voice information.

[0310] "Means for transmitting generated audio data" refers to devices or software that have the function of transmitting audio data generated by a speech synthesis engine to a designated terminal or receiving device.

[0311] "Means for passing text data and audio data to emotion recognition means and analyzing the user's emotions" refers to devices or software that have the function of determining emotions from text and audio using an analysis engine and identifying the speaker's emotional state.

[0312] "Means of adjusting response content based on emotions" refers to devices or software that have the function of modifying responses to appropriate tone and wording based on emotional information identified by an analysis engine.

[0313] "Means of notifying a user when specific keywords are detected, allowing the user to intervene in a call" refers to devices or software that have the function of notifying an administrator or operator when pre-set important keywords or emotional states are detected, providing them with an opportunity to intervene in the situation.

[0314] The system related to this invention automates telephone answering, allowing user intervention as needed. Furthermore, the system incorporates an emotion engine that recognizes the user's emotions and adjusts its response accordingly. This system includes the following elements:

[0315] 1. Terminal

[0316] 2. Server

[0317] 3. User

[0318] 4. Emotional Engine

[0319] Receiving and converting audio data

[0320] The device detects an incoming call and, once the call begins, sends audio data to the server in real time. This audio data, containing the conversation content, is transmitted to the server continuously. Specifically, smartphones and landline phones fulfill this role.

[0321] Text data analysis and response generation

[0322] The server receives the audio data and converts it into text data in real time using a speech recognition engine (e.g., Google Cloud Speech-to-Text API). The converted text data is then passed to a natural language processing engine (e.g., Google Cloud Natural Language API) to analyze the content and intent of the conversation.

[0323] Next, based on the analysis results, a generative AI (e.g., OpenAI®'s GPT-4®) is used to generate an appropriate response in text format. The generated text response is then passed to a speech synthesis engine (e.g., Microsoft Azure Text-to-Speech) and converted into speech data in real time.

[0324] Emotion recognition and response adjustment by an emotion engine

[0325] The server passes text and audio data to an emotion engine (e.g., IBM Watson® Tone Analyzer) to analyze the user's emotions. Based on the analysis, the response is adjusted. For example, if the user is angry, the response will be more careful and polite.

[0326] Sending a response and notification

[0327] The server sends the generated audio data to the terminal, which then sends it back to the caller. Simultaneously, if important call content or emotions are detected, the user is notified. The notification might say something like, "We are handling a complaint. Intervention is required."

[0328] User intervention

[0329] The device displays a notification to the user and offers an option to intervene. If the user chooses to intervene, the call is seamlessly handed over to the user. The user can then acknowledge the notification and respond directly.

[0330] Specific example

[0331] For example, consider a scenario where a customer calls to complain, saying, "The item I ordered hasn't arrived yet!"

[0332] The device sends the audio data to the server.

[0333] The server converts the audio data into text and analyzes it using an NLP engine.

[0334] The server detects keywords such as "order," "not received," and "dissatisfaction," and uses a generation AI to generate a response such as "We apologize, we will check on this immediately."

[0335] The emotion engine detects an angry tone and carefully adjusts its response accordingly.

[0336] The server sends the response to the customer via the terminal.

[0337] Based on sentiment data and keywords, the server sends a notification to the user stating, "We are handling a complaint; intervention is required."

[0338] The user reviews the notification and chooses to intervene. The system responds directly to the customer with, "We will investigate immediately, please wait."

[0339] Example of a prompt

[0340] "We have received an inquiry from a customer regarding a non-delivery of their product, and their message is filled with anger. Please prepare an appropriate response."

[0341] "The customer is dissatisfied with the delay in their order. Please generate a polite response to improve this situation."

[0342] In this way, the system of the present invention incorporates emotion recognition functionality to provide appropriate responses based on the user's emotions, and allows the user to intervene as needed. This enables improved efficiency and quality in telephone communication.

[0343] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0344] Step 1:

[0345] The device detects incoming calls and, once a call begins, sends audio data to the server in real time. Specifically, the device's phone function detects the start of a call, converts the audio picked up by the microphone into digital audio data, and sends it to the server via the internet.

[0346] Input: Incoming phone call

[0347] Output: Audio data transmitted in real time

[0348] Step 2:

[0349] The server receives audio data and converts it into text data in real time using a speech recognition engine (e.g., Google Cloud Speech-to-Text API). Specifically, the speech recognition engine analyzes the audio signal and converts the words used in the conversation into text format.

[0350] Input: Audio data transmitted in real time

[0351] Output: Text data

[0352] Step 3:

[0353] The server passes the converted text data to a natural language processing engine (e.g., Google Cloud Natural Language API) to analyze the conversation. Specifically, the natural language processing engine extracts meaning from the text data and identifies important keywords and topics.

[0354] Input: Text data converted by the speech recognition engine

[0355] Output: Analyzed keyword and topic information

[0356] Step 4:

[0357] The server uses a generative AI model (e.g., OpenAI's GPT-4) based on the analysis results to generate an appropriate response in text format. It utilizes prompts to adjust the generated response as needed.

[0358] Input: Analyzed keyword and topic information

[0359] Output: Generated text response

[0360] Example prompt:

[0361] "We have received an inquiry from a customer regarding a non-delivery of their product, and their message is filled with anger. Please prepare an appropriate response."

[0362] Step 5:

[0363] The server passes the generated text response to a text-to-speech engine (e.g., Microsoft Azure Text-to-Speech), which converts it into speech data in real time. Specifically, the text-to-speech engine converts the text into speech, creating natural and fluent speech output.

[0364] Input: Generated text response

[0365] Output: Generated audio data

[0366] Step 6:

[0367] The server passes the voice and text data to an emotion recognition engine (e.g., IBM Watson Tone Analyzer) to analyze the user's emotions. Specifically, the emotion recognition engine identifies emotions from the tone and word choice of the voice and then identifies areas that need adjustment based on that.

[0368] Input: Generated text response and audio data

[0369] Output: Emotion analysis results

[0370] Step 7:

[0371] The server adjusts the generated response based on the recognized emotion. For example, if anger is detected, the response will be made more careful and polite.

[0372] Input: Sentiment analysis results

[0373] Output: Adjusted text response and audio data

[0374] Step 8:

[0375] The server sends the generated audio data to the terminal, and the terminal sends it back to the person on the other end of the line. Specifically, the server sends audio data to the terminal in real time, and the terminal transmits that data to the person on the other end of the line through its audio output device (speaker).

[0376] Input: Adjusted audio data

[0377] Output: Transmitted audio data

[0378] Step 9:

[0379] Based on text data and sentiment analysis results, the server sends notifications to the user when specific keywords or important emotions are detected. These notifications may include alerts such as, "We are handling a complaint. Intervention is required."

[0380] Input: Specific keywords and sentiment analysis results

[0381] Output: User statement notification

[0382] Step 10:

[0383] The device displays a notification to the user and offers an option to intervene. If the user chooses to intervene, the call is seamlessly handed over to the user. Specifically, the user acknowledges the notification and responds directly to the call if necessary.

[0384] Input: User Statement Notification

[0385] Output: User intervention to handle calls

[0386] (Application Example 2)

[0387] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as a "server" and the smart device 14 as a "terminal".

[0388] Traditional telephone answering systems failed to consider customer emotions, and inappropriate responses easily led to customer dissatisfaction. Furthermore, automated answering systems could only handle a limited range of questions, requiring human intervention for complex inquiries and complaints, and this intervention was sometimes not timely. In addition, content delivery services require prompt and appropriate responses to inquiries to improve the user experience.

[0389] The specific processing performed by the specific 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 receiving voice data and converting it into text data in real time, means for analyzing the text data and generating a response using a generation AI model based on the analysis results, means for converting the generated response into voice data, means for analyzing the user's emotions from the text data and voice data using an emotion engine, means for transmitting the voice data of the generated response, and means for notifying the user when specific emotions and keywords are detected and providing an option for intervention. This provides an appropriate automated response that takes the user's emotions into consideration, and allows for human intervention at the appropriate time as needed.

[0390] "Audio data" refers to information that represents telephone calls or recorded audio in digital format.

[0391] "Text data" refers to character information generated by analyzing audio data and using speech recognition technology.

[0392] A "generative AI model" is a type of artificial intelligence that automatically generates appropriate responses to specific inputs (prompt sentences) based on natural language processing.

[0393] A "prompt statement" is an input statement used to generate a response for a generative AI model.

[0394] An "emotion engine" is an algorithm or system that analyzes voice and text data to recognize the user's emotions (such as joy, anger, or sadness) contained within it.

[0395] "Analysis results" refer to conclusions or information obtained through analysis of text data using natural language processing or sentiment engines.

[0396] "Response" refers to the reply or action generated by a generative AI model based on its analysis results.

[0397] A "notification" is an alert or message sent from a system to a user under specific conditions (for example, detection of the user's emotions or keywords).

[0398] "Intervention" refers to the act of a user directly participating in a call or response when the system cannot handle the situation through automated responses or under specific conditions.

[0399] "Means" refer to equipment, algorithms, systems, methods, etc., used to achieve a specific purpose.

[0400] The present invention relates to a system that automates telephone answering and recognizes user emotions to adjust responses. The system comprises means for receiving voice data and converting it into text data in real time, means for analyzing the text data and generating responses using a generative AI model, means for converting the generated responses into voice data, means for analyzing the user's emotions using an emotion engine, means for transmitting the generated voice data, and means for notifying the user when specific emotions or keywords are detected and providing options for intervention.

[0401] System Overview

[0402] Receiving and converting audio data:

[0403] The device detects an incoming call and sends audio data to the server in real time when the call begins. The server uses a speech recognition engine (e.g., Google Speech Recognition) to convert the audio data into text data.

[0404] Text data analysis and response generation:

[0405] The server passes text data to an NLP engine, which analyzes the content and intent of the conversation. Based on the analysis results, it generates a response using a generative AI model (e.g., the OpenAI GPT model). At this time, prompts are used to guide the model in generating the response.

[0406] Emotion recognition by an emotion engine:

[0407] Text and audio data are passed to an emotion engine (e.g., Hugging Face's sentiment-analysis pipeline) to analyze the user's emotions. Based on the recognized emotions, the server adjusts its response.

[0408] Sending responses and notifications:

[0409] The generated audio data is sent to the terminal and then forwarded from the terminal to the customer. Simultaneously, the server detects specific emotions and keywords, and if important call content or emotions are detected, it notifies the user. The user reviews the notification, and if they choose to intervene, the call is seamlessly handed over to them.

[0410] Specific example

[0411] For example, if a customer says, "The video won't play," the system will process it as follows:

[0412] The terminal sends audio data to the server, and the server converts the audio data into text.

[0413] The system analyzes text data and uses a generative AI model to generate an appropriate response ("We apologize, we will investigate").

[0414] The emotion engine detects user dissatisfaction, and the server adjusts its response to be more attentive.

[0415] The server converts the response into audio data and sends it to the customer from the terminal.

[0416] When emotions and keywords are detected, a notification is sent to the user, and they are given the option to intervene.

[0417] Example of a prompt:

[0418] A user said: "The video won't play." How do you respond?

[0419] The system described above enables prompt and appropriate responses to customer inquiries and complaints. Furthermore, by providing responses that take user emotions into account, improved customer satisfaction can be expected.

[0420] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0421] Step 1:

[0422] The terminal detects an incoming call and receives audio data in real time when the call begins. It then sends the received audio data to the server. Specifically, the terminal uses a microphone to capture audio and transfers the audio data to the server in digital format. In this step, the input is the audio data, and the output is the audio data sent to the server.

[0423] Step 2:

[0424] The server converts the received audio data into text data in real time using a speech recognition engine (e.g., Google Speech Recognition). The server converts the audio to text using digital signal processing and further generates text data for the next processing step. In this step, the input is audio data and the output is text data.

[0425] Step 3:

[0426] The server passes text data to a natural language processing engine for analysis. During the analysis process, it extracts specific intentions and themes and generates data to provide to the AI ​​model that generates appropriate prompt sentences. For example, the keyword "cannot be played" may be detected from the text data. In this step, the input is text data, and the output is the analysis results and prompt sentences for the AI ​​model.

[0427] Step 4:

[0428] The server inputs a prompt message into the generative AI model, which then generates an appropriate response. The generative AI model generates a text-based response based on the provided prompt message. This text-based response contains the solution to the user's question or problem. In this step, the input is the prompt message, and the output is the text data of the generated response.

[0429] Step 5:

[0430] The server passes the generated response text data to a speech synthesis engine (e.g., pyttsx3) to convert it into speech data. The speech synthesis engine converts the text into natural-sounding speech, making it ready to send the response to the customer. In this step, the input is the response text data, and the output is the response speech data.

[0431] Step 6:

[0432] The server passes text and audio data to the emotion engine (e.g., Hugging Face's sentiment-analysis pipeline) to analyze the user's emotions. This analysis identifies the user's positive, negative, or neutral emotions and may reflect them in the response. In this step, the input is text and audio data, and the output is the emotion analysis result.

[0433] Step 7:

[0434] The server further refines the generated response based on the sentiment analysis results, providing it in a way that is appropriate to the user's emotions. For example, if the user is angry, the response is adjusted to be more polite. In this step, the input is the sentiment analysis results and the generated response text data, and the output is the adjusted response text data.

[0435] Step 8:

[0436] The server sends the finalized response audio data to the terminal, which then sends it back to the caller. The terminal plays the audio back to the customer in real time, conveying the content of the response. In this step, the input is the adjusted response audio data, and the output is the response to the customer.

[0437] Step 9:

[0438] If specific emotions and keywords are detected, the server sends a notification to the user informing them that intervention is needed. The user reviews the notification and chooses whether to intervene in the call if necessary. In this step, the inputs are the emotion analysis results and keyword detection results, and the outputs are the notification to the user and the choice of intervention.

[0439] 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 user input for the result of the specific processing. The control unit 46A transmits the audio data indicating 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.

[0440] Data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of data generation model 58 is ChatGPT (registered trademark) (Internet search).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0441] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart device 14.

[0442] [Second Embodiment]

[0443] Figure 3 shows an example of the configuration of the data processing system 210 according to the second embodiment.

[0444] As shown in Figure 3, the data processing system 210 includes a data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0445] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0446] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication interface 44. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, and camera 42 are also connected to the bus 52.

[0447] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0448] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0449] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0450] Figure 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Figure 4, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0451] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[0452] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0453] In the smart glasses 214, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0454] Next, the identification processing performed by the identification processing unit 290 of the data processing device 12 will be described. 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".

[0455] The system according to the present invention receives voice data, converts it to text data in real time, analyzes it, generates an appropriate response, and transmits it as voice data. It also detects specific keywords and notifies the user, enabling seamless switching so that the user can intervene as needed.

[0456] System-wide configuration

[0457] The system consists of the following elements:

[0458] 1. Terminal

[0459] 2. Server

[0460] 3. User

[0461] Receiving and converting audio data

[0462] The device detects an incoming call and, as soon as the call begins, sends the audio data to the server in real time.

[0463] The server passes the received audio data to the speech recognition engine, which converts it into text data in real time.

[0464] Text data analysis and response generation

[0465] The server uses a natural language processing (NLP) engine to analyze text data and understand its content.

[0466] Based on the analysis results, the server automatically generates an appropriate response using a generation AI. This response is then generated again in text format.

[0467] The server passes the text-based response to the speech synthesis engine, which converts it into speech data.

[0468] Sending a response and notification

[0469] The server sends the generated voice data to the terminal, which then relays it to the person on the other end of the line.

[0470] The server continuously monitors the call content and notifies the user via the terminal if it detects specific keywords.

[0471] Seamless user intervention

[0472] Users can check notifications and, if important call content is detected, intervene in the call themselves.

[0473] The device accepts user intervention, allowing the user to speak directly with the person they are calling.

[0474] Specific example

[0475] 1. Handling sales calls

[0476] A salesman makes a phone call and says, "Excuse me for bothering you while you're busy, this is a salesperson from AA Insurance."

[0477] The device transmits voice data to the server in real time.

[0478] The server converts the voice data into text and recognizes it as "This is a salesperson from AA Insurance."

[0479] The server uses a generation AI to generate the response, "I'm not interested, thank you," converts it into audio data, and sends it to the terminal.

[0480] The device responds to the salesman by playing this audio data.

[0481] 2. Important notifications and user intervention

[0482] The salesman continued, "Would you mind if I sent you some materials?"

[0483] The server detects specific keywords such as "send documents" and notifies the user via the terminal.

[0484] The user checks the notification and taps the screen to intervene in the call, directly responding, "I don't need to send any documents."

[0485] Thus, the present invention is a system that automates telephone answering while allowing user intervention in important situations, thereby achieving efficient and flexible responses.

[0486] Next, we can describe the details of the processing steps.

[0487] The following describes the processing flow.

[0488] Step 1:

[0489] The device detects an incoming call and collects the call audio as soon as the call begins. The collected audio data is transmitted to the server in real time.

[0490] Step 2:

[0491] The server passes the received audio data to the speech recognition engine, which converts the audio data into text data. This conversion is performed in real time and processed sequentially.

[0492] Step 3:

[0493] The server passes the text data to a natural language processing (NLP) engine for content analysis. The analysis identifies the purpose and topic of the call.

[0494] Step 4:

[0495] The server then uses the generation AI to create a response based on the analysis results. This response is generated in text format and designed to be an appropriate reply to the conversation.

[0496] Step 5:

[0497] The server passes the generated text response to the speech synthesis engine, which converts it into speech data. The converted speech data is then generated again in real time.

[0498] Step 6:

[0499] The server sends the audio data to the terminal. The terminal decodes the received audio data and plays it back to the caller at the appropriate time.

[0500] Step 7:

[0501] The server continues to monitor the call content to check for the inclusion of specific keywords or phrases. For example, it may detect specific keywords such as "important matter" or "sending documents."

[0502] Step 8:

[0503] When a specific keyword is detected, the server sends a notification to the terminal, informing the user that intervention is necessary.

[0504] Step 9:

[0505] The device displays a notification to the user, alerting them with audio and visuals. It then prompts the user to choose whether or not to take action.

[0506] Step 10:

[0507] The user checks the notification and, if important call content is detected, intervenes in the call at their own discretion. The device accepts the user's intervention and allows the user to speak directly with the other party.

[0508] Step 11:

[0509] Once the call ends, the device sends the audio recording and text data to the server. The server stores this data in a database for later reference.

[0510] In this way, the system seamlessly integrates real-time voice responses with user intervention, enabling effective telephone handling.

[0511] (Example 1)

[0512] Next, we will describe Example 1. 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".

[0513] Conventional voice response systems were unable to adequately generate responses or detect keywords in real time, making it difficult to respond immediately to important call content. Furthermore, the inability to seamlessly integrate automated responses with user intervention made flexible and efficient call handling challenging.

[0514] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0515] In this invention, the server includes means for receiving voice data and converting it into text data in real time; means for analyzing the text data and generating an appropriate response using a generative AI model; means for converting the generated text data of the response into voice data; means for transmitting the generated voice data of the response; means for detecting specific keywords and notifying the user; and means for allowing the user to acknowledge the notification and intervene in the call. This enables real-time text conversion and response generation, keyword detection and user notification, and seamless user intervention.

[0516] "Audio data" refers to audio signals obtained through phone calls, recordings, etc., represented in digital format.

[0517] "Real-time" refers to data processing and responses occurring immediately, with results reflected without delay.

[0518] "Text data" refers to audio data converted into written information, expressed in a format that is understandable to humans.

[0519] A "generative AI model" refers to a technology that uses artificial intelligence algorithms to generate appropriate responses or data based on given input.

[0520] A "prompt" is a series of text inputs to a generative AI model, which the AI ​​uses to generate a response.

[0521] "Analysis" is a means of processing text data to understand its content and structure, and a process for deriving appropriate responses or actions.

[0522] A "keyword" is an important word or phrase contained within text data, and its detection triggers a specific action or response.

[0523] A "user" is a person who uses the system and has the authority to intervene in the content of a call.

[0524] A "notification" is an informational alert sent from the system to the user, intended to draw the user's attention when a specific keyword is detected.

[0525] "Intervening in a call" refers to the action of a user switching from an automated system response to a manual response and starting a direct conversation with the other party.

[0526] A "speech recognition engine" is software or hardware that analyzes speech data and converts it into text data.

[0527] A "speech synthesis engine" is software or hardware that analyzes text data and converts it into speech data.

[0528] "Communication methods" refer to technologies used to send and receive voice and text data between a server and a terminal, and include the Internet Protocol and other communication protocols.

[0529] A specific embodiment of the system according to the present invention is shown below. This system receives voice data, converts it into text data in real time, performs subsequent analysis and response generation, and enables notifications and call interventions to the user.

[0530] Receiving and converting audio data

[0531] The terminal detects an incoming call and captures audio data as soon as the call begins. The call audio is transmitted to the server in real time. WebSocket and HTTP streaming technologies are used as secure communication methods for transmitting the audio data.

[0532] The server converts the received audio data into text data using the Google Cloud Speech-to-Text API or Amazon Transcribe. The converted text data is extracted as a string that accurately represents the spoken content.

[0533] Text data analysis and response generation

[0534] The server uses Google Cloud Natural Language API and Microsoft Azure Text Analytics to analyze text data and understand its content. The information extracted through analysis includes grammatical structure, meaning, and important keywords.

[0535] Based on the analysis results, the server automatically generates an appropriate response using a generative AI model (e.g., GPT-3). The generative AI model used here has advanced natural language generation capabilities and generates the optimal response from the given prompt sentence.

[0536] The generated response is then converted back into text format and then into speech data using the Google Text-to-Speech API or Amazon Polly.

[0537] Sending a response and notification

[0538] The server sends the generated audio data to the terminal, which then relays that audio data to the person on the other end of the call. This enables real-time responses.

[0539] A server that continuously monitors call content will notify the user via the terminal if it detects a specific keyword, such as the phrase "send documents." The notification system can display the notification on the terminal screen or alert the user with vibration or sound.

[0540] Seamless user intervention

[0541] Upon receiving a notification, the user can review the notification displayed on their device screen and decide whether or not to intervene in the call. If the user chooses to intervene, tapping the notification on their device will switch the system from automatic response to manual response, allowing the user to speak directly with the caller.

[0542] Specific example

[0543] 1. Handling sales calls

[0544] A salesman makes a phone call and says, "Excuse me for bothering you while you're busy, this is a salesperson from AA Insurance."

[0545] The device transmits voice data to the server in real time.

[0546] The server converts the voice data into text and recognizes it as "This is a salesperson from AA Insurance."

[0547] The server uses a generation AI to generate the response, "I'm not interested, thank you," converts it into audio data, and sends it to the terminal.

[0548] The device responds to the salesman by playing this audio data.

[0549] 2. Important notifications and user intervention

[0550] The salesman continued, "Would you mind if I sent you some materials?"

[0551] The server detects specific keywords such as "send documents" and notifies the user via the terminal.

[0552] The user checks the notification, taps the screen to intervene in the call, and directly responds, "I don't need to send any documents."

[0553] Example of a prompt

[0554] "If a salesperson is soliciting insurance and includes the phrase 'send me some materials,' the user should be notified. The response should generate 'I do not need any materials.'"

[0555] This system enables efficient and flexible call handling by transcribing audio in real time, generating appropriate responses, and allowing user intervention at key points.

[0556] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0557] Step 1:

[0558] The device detects an incoming call and captures audio data as soon as the call begins. The audio data is sent to the server in real time. Specifically, it receives the phone's audio in real time as input and sends that audio data to the server via the network as output. In practice, the smartphone outputs a log message saying "Audio capture started" and immediately begins sending the audio data.

[0559] Step 2:

[0560] The server passes the received audio data to a speech recognition engine (e.g., Google Cloud Speech-to-Text API) and converts it into text data in real time. It takes audio data as input and generates text data as output. Specifically, the server outputs a log message saying "Audio data converted to text."

[0561] Step 3:

[0562] The server uses a natural language processing (NLP) engine (e.g., Google Cloud Natural Language API) to analyze text data. It receives text data as input and generates analysis results as output. Analysis includes entity extraction and sentiment analysis. Specifically, the server logs "Text analysis complete: Intent is sales."

[0563] Step 4:

[0564] The server generates an appropriate response based on the analysis results using a generative AI model (e.g., GPT-3). It receives the analysis results and a prompt as input and generates the appropriate response text as output. Specifically, the server logs "Response generation complete." An example of a prompt is: "If a salesman is soliciting insurance and includes the phrase 'send materials,' the user should be notified. The response should generate 'No materials are needed.'"

[0565] Step 5:

[0566] The server passes the generated response text to a text-to-speech engine (e.g., Google Text-to-Speech API) to convert it into speech data. It takes response text as input and generates speech data as output. Specifically, the server logs "Text-to-speech complete".

[0567] Step 6:

[0568] The server sends the generated audio data to the terminal, and the terminal transmits that audio data to the person it is talking to. It receives audio data as input and plays it back to the person it is talking to as output. Specifically, the terminal log displays "Playing response audio".

[0569] Step 7:

[0570] The server continuously monitors the call content and, if it detects a specific keyword (e.g., "send documents"), it notifies the user via the terminal. It receives real-time text data as input and generates a notification as output. Specifically, the server log records "Keyword detected: Send documents" and sends a notification to the terminal.

[0571] Step 8:

[0572] The user can check notifications and, if important call content is detected, intervene in the call themselves. The device receives notifications as input and intervenes in the call as output. Specifically, the user taps the notification on their device and directly responds with "No documents are needed." The device then logs "User intervened in the call."

[0573] (Application Example 1)

[0574] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0575] Traditional content distribution services have faced challenges in providing personalized recommendations based on user preferences and interests, making it difficult to respond quickly and accurately to the information and entertainment users seek. Furthermore, the lack of systems that use AI-generated models to respond to requests has made it difficult to obtain the information users need in real time.

[0576] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0577] In this invention, the server includes means for receiving voice data and converting it into text data in real time; means for analyzing the text data, generating a response using a generative AI model based on the analysis results, and converting the response into voice data; means for transmitting the generated voice data of the response; means for obtaining a request from the user in voice data and sending a prompt sentence to the generative AI model based on the request; and means for suggesting appropriate content to the user based on the generated prompt. This makes it possible to recommend the most suitable content to the user in real time.

[0578] "Audio data" refers to data obtained by converting audio into a digital format.

[0579] "Real-time" refers to processing data instantly and without delay.

[0580] "Text data" refers to digital data obtained by converting audio data into written text.

[0581] A "generative AI model" refers to an artificial intelligence model that generates appropriate responses based on user input data.

[0582] A "prompt" is a portion of the text data input to a generating AI model, and it functions as an instruction or question for the model.

[0583] "Analysis" refers to the process of understanding text data and extracting its meaning.

[0584] "Content" is a general term for information and entertainment that users consume, such as movies, music, and news.

[0585] "Suggestions" refer to the act of recommending appropriate content based on the user's preferences and requests.

[0586] This invention relates to a system that converts audio data into text data in real time, analyzes the text data to generate a response using a generative AI model, and transmits that response back as audio data. Furthermore, it also includes a function to generate prompt sentences for the generative AI model based on user requests and suggest appropriate content.

[0587] System Configuration

[0588] The system consists of the following elements:

[0589] 1. Terminal

[0590] An input device for receiving audio data (e.g., a microphone)

[0591] An output device for playing audio data (e.g., a speaker)

[0592] 2. Server

[0593] Speech recognition engine

[0594] Natural Language Processing (NLP) Engine

[0595] Generative AI models (e.g., GPT-3)

[0596] Speech synthesis engine

[0597] 3. User

[0598] Person making a request by voice

[0599] Receiving and converting audio data

[0600] The terminal receives the user's voice data and sends it to the server in real time. The server uses a speech recognition engine to convert the voice data into text data.

[0601] Text data analysis and response generation

[0602] The server analyzes the converted text data using a natural language processing (NLP) engine to understand its content. Based on the analysis results, it uses a generative AI model to generate an appropriate response, which is then generated again in text format. The generated text is then converted into speech data by a speech synthesis engine.

[0603] Sending a response and suggesting content

[0604] The generated audio data is sent to the device and transmitted to the user through the device's speaker. Furthermore, a request is obtained from the user via the audio data, and a prompt is sent to the generating AI model based on that request. Based on the generated prompt, appropriate content is suggested to the user.

[0605] Specific example

[0606] For example, if a user requests "Tell me your recommendations for the latest movies," the following prompt message will be generated.

[0607] Examples of prompts to input into a generative AI model:

[0608] User request: Please recommend some of the latest movies.

[0609] Appropriate response:

[0610] Based on this prompt, the server sends information to the AI ​​model that generates the appropriate response text. For example, a response such as "As a recent movie recommendation, we suggest the action film XXX" is generated, converted into audio data, and delivered to the user.

[0611] This allows users to receive real-time responses to their requests and to have content that interests them suggested quickly and accurately.

[0612] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0613] Step 1:

[0614] The device receives the user's voice data.

[0615] Input: User's voice data

[0616] Operation: The microphone captures the user's voice and converts the audio data into a digital format.

[0617] Output: Digital audio data

[0618] Step 2:

[0619] The terminal sends the received digital audio data to the server.

[0620] Input: Digital audio data

[0621] Operation: Transmits audio data to the server in real time via the network interface.

[0622] Output: Audio data sent to the server

[0623] Step 3:

[0624] The server uses a speech recognition engine to convert the audio data into text data.

[0625] Input: Audio data

[0626] Operation: Analyzes audio data using a speech recognition engine (e.g., Google Speech Recognition) and converts it to text.

[0627] Output: Text data

[0628] Step 4:

[0629] The server analyzes the text data using a natural language processing (NLP) engine and understands its meaning.

[0630] Input: Text data

[0631] Operation: Uses a natural language processing (NLP) engine to extract meaning from text data.

[0632] Output: Analysis results (user request details)

[0633] Step 5:

[0634] Based on the analysis results, the server generates a response using an AI model.

[0635] Input: Analysis results

[0636] Operation: A prompt is sent to a generative AI model (e.g., GPT-3), and a response text is generated based on it. An example prompt is "User request: Tell me your recommendations for the latest movies. Appropriate response:"

[0637] Output: Response text data

[0638] Step 6:

[0639] The server converts the generated response text into speech data using a speech synthesis engine.

[0640] Input: Response text data

[0641] Operation: Uses a speech synthesis engine (e.g., pyttsx3) to convert response text into speech data.

[0642] Output: Response voice data

[0643] Step 7:

[0644] The server sends the generated response voice data to the terminal.

[0645] Input: Response voice data

[0646] Operation: Transmits voice data to the terminal in real time via the network interface.

[0647] Output: Response voice data sent to the terminal

[0648] Step 8:

[0649] The device plays back the received voice response data and conveys it to the user.

[0650] Input: Response voice data

[0651] Operation: Plays audio data through the speaker.

[0652] Output: The user receives a voice response.

[0653] Step 9:

[0654] If a new request comes from the user, the same process is repeated.

[0655] Input: New voice request

[0656] Action: Repeat the process from step 1.

[0657] Output: New response to the user

[0658] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0659] The system related to this invention automates telephone answering and allows for seamless user intervention as needed. Furthermore, this system incorporates an emotion engine that recognizes the user's emotions and adjusts its response accordingly.

[0660] System-wide configuration

[0661] The system consists of the following elements:

[0662] 1. Terminal

[0663] 2. Server

[0664] 3. User

[0665] 4. Emotional Engine

[0666] Receiving and converting audio data

[0667] The device detects an incoming call and sends audio data to the server in real time when the call begins. The audio data is the content of the call and is sent to the server sequentially.

[0668] Text data analysis and response generation

[0669] The server receives the audio data and converts it into text data in real time using a speech recognition engine.

[0670] The server passes the text data to the NLP engine, which analyzes the content and intent of the conversation. This clarifies the specific topic and purpose of the call.

[0671] Based on the analysis results, the server uses a generative AI to generate an appropriate response in text format. This process builds an automated response system for calls.

[0672] The server passes the generated text response to the speech synthesis engine, which converts it into speech data in real time.

[0673] Emotion recognition by an emotion engine

[0674] The server passes text and audio data to the emotion engine, which analyzes the user's emotions. For example, emotions can be inferred from the tone of voice and word choice.

[0675] Based on the emotions recognized by the emotion engine, the server adjusts its response. If the user is angry, a more careful response will be generated.

[0676] The server detects specific emotions and keywords and sends a notification to the user based on them. In this case, it is determined that the user needs to intervene.

[0677] Sending a response and notification

[0678] The server sends the generated audio data to the terminal, which then sends it back to the person on the other end of the call. Simultaneously, if important conversation content or emotions are detected, the user is notified.

[0679] The device displays a notification to the user and offers them an option to intervene. If the user chooses to intervene, the call is seamlessly handed over to the user.

[0680] Specific example

[0681] 1. Examples of handling customer complaints

[0682] A customer calls to complain: "The item I ordered hasn't arrived yet!"

[0683] The device sends the audio data to the server.

[0684] The server converts the audio data into text and analyzes it using an NLP engine.

[0685] The server detects keywords such as "order," "not received," and "dissatisfaction," and uses a generation AI to generate a response such as "We apologize, we will check on this immediately."

[0686] The emotion engine detects an angry tone and carefully adjusts its response accordingly.

[0687] The server sends the response to the customer via the terminal.

[0688] The server sends a notification to the user based on sentiment data and keywords. It displays a message saying, "We are handling a complaint and intervention is required."

[0689] The user reviews the notification and selects intervention. The system responds directly to the customer with, "We will investigate immediately, please wait."

[0690] In this way, the system of the present invention incorporates emotion recognition functionality to provide appropriate responses based on the user's emotions, and allows the user to intervene as needed. This enables improved efficiency and quality in telephone communication.

[0691] The following describes the processing flow.

[0692] Step 1:

[0693] The device detects incoming calls and collects call audio as soon as the call begins. The collected audio data is sent to the server in real time.

[0694] Step 2:

[0695] The server passes the received audio data to the speech recognition engine, which converts the audio data into text data. This conversion is performed in real time and processed sequentially.

[0696] Step 3:

[0697] The server then passes the text data to a natural language processing (NLP) engine for analysis. The analysis results identify the purpose and topic of the call.

[0698] Step 4:

[0699] The server passes the analyzed text data to the emotion engine, which analyzes the user's emotions. The emotion engine identifies the user's emotions based on their tone of voice and word choice.

[0700] Step 5:

[0701] The server integrates the results from the emotion engine and the NLP engine's analysis, and has the generative AI create an appropriate response. This response is generated in text format and is designed to be an appropriate reply in a conversation.

[0702] Step 6:

[0703] The server passes the generated text response to the speech synthesis engine, which converts it into speech data. The generated speech data is then generated again in real time.

[0704] Step 7:

[0705] The server sends the generated voice data to the terminal, which then sends it to the caller. This is where the response is conveyed to the other party.

[0706] Step 8:

[0707] The server continues to monitor the call content and sends notifications to the user if specific keywords or emotions are detected.

[0708] Step 9:

[0709] The device displays notifications to the user, alerting them with audio and visuals. These notifications may include messages such as, "A customer is expressing dissatisfaction. Intervention is needed."

[0710] Step 10:

[0711] The user reviews the notification and, if important call content or emotions are detected, intervenes in the call at their own discretion. The device accepts the user's intervention.

[0712] Step 11:

[0713] The user chooses to intervene and begins a direct conversation with the person on the other end of the line. This allows for a seamless integration of human judgment into responses, even when emotional responses are required.

[0714] Step 12:

[0715] Once the call ends, the device sends the audio recording and text data to the server. The server stores this data in a database for later reference.

[0716] By combining this with an emotion engine, it becomes possible to automate telephone answering while still providing flexible responses based on the user's emotions.

[0717] (Example 2)

[0718] Next, we will describe Example 2. 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".

[0719] Traditional telephone answering systems often relied on manual processes for analyzing call content and generating appropriate responses, resulting in challenges in response quality and efficiency. Furthermore, they struggled to respond in a way that considered user emotions, hindering responses that required quick and appropriate action, particularly in situations like complaint handling. Additionally, delays in detecting specific key keywords or emotions could lead to delayed user intervention and inadequate responses.

[0720] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0721] In this invention, the server includes means for detecting an incoming call and transmitting voice data in real time when the call begins; means for receiving the voice data and converting it into text data in real time using recognition means; means for passing the text data to analysis means and creating a response based on the analysis results using generation means; means for passing the text data of the generated response to synthesis means and converting it into voice data; and means for transmitting the generated voice data. This enables real-time analysis of call content and automatic generation of appropriate responses.

[0722] "Means for detecting incoming phone calls" refers to devices or software that identify call requests received via telephone lines or internet communication.

[0723] "Means of transmitting voice data in real time" refers to devices or software that have the function of instantly processing voice information spoken during a call and transmitting it to other devices or systems via a network.

[0724] "Means for receiving audio data and converting it into text data in real time using recognition means" refers to devices or software that have the function of instantly converting received audio information into text data using a speech recognition engine.

[0725] "Means for passing text data to an analysis means and creating a response based on the analysis results using a generation means" refers to devices or software that send text data to a natural language processing engine and create an appropriate response sentence using a generation AI or similar based on the analysis results.

[0726] "Means for passing the generated response text data to a synthesis means and converting it into audio data" refers to devices or software that have the function of sending a text-based response to a speech synthesis engine and converting it into a format that can be played back as audio information.

[0727] "Means for transmitting generated audio data" refers to devices or software that have the function of transmitting audio data generated by a speech synthesis engine to a designated terminal or receiving device.

[0728] "Means for passing text data and audio data to emotion recognition means and analyzing the user's emotions" refers to devices or software that have the function of determining emotions from text and audio using an analysis engine and identifying the speaker's emotional state.

[0729] "Means of adjusting response content based on emotions" refers to devices or software that have the function of modifying responses to appropriate tone and wording based on emotional information identified by an analysis engine.

[0730] "Means of notifying a user when specific keywords are detected, allowing the user to intervene in a call" refers to devices or software that have the function of notifying an administrator or operator when pre-set important keywords or emotional states are detected, providing them with an opportunity to intervene in the situation.

[0731] The system related to this invention automates telephone answering, allowing user intervention as needed. Furthermore, the system incorporates an emotion engine that recognizes the user's emotions and adjusts its response accordingly. This system includes the following elements:

[0732] 1. Terminal

[0733] 2. Server

[0734] 3. User

[0735] 4. Emotional Engine

[0736] Receiving and converting audio data

[0737] The device detects an incoming call and, once the call begins, sends audio data to the server in real time. This audio data, containing the conversation content, is transmitted to the server continuously. Specifically, smartphones and landline phones fulfill this role.

[0738] Text data analysis and response generation

[0739] The server receives the audio data and converts it into text data in real time using a speech recognition engine (e.g., Google Cloud Speech-to-Text API). The converted text data is then passed to a natural language processing engine (e.g., Google Cloud Natural Language API) to analyze the content and intent of the conversation.

[0740] Next, based on the analysis results, a generative AI (e.g., OpenAI's GPT-4) is used to generate an appropriate response in text format. The generated text response is then passed to a speech synthesis engine (e.g., Microsoft Azure Text-to-Speech) and converted into speech data in real time.

[0741] Emotion recognition and response adjustment by an emotion engine

[0742] The server passes text and audio data to an emotion engine (e.g., IBM Watson Tone Analyzer) to analyze the user's emotions. Based on the analysis, the response is adjusted. For example, if the user is angry, the response will be more careful and polite.

[0743] Sending a response and notification

[0744] The server sends the generated audio data to the terminal, which then sends it back to the caller. Simultaneously, if important call content or emotions are detected, the user is notified. The notification might say something like, "We are handling a complaint. Intervention is required."

[0745] User intervention

[0746] The device displays a notification to the user and offers an option to intervene. If the user chooses to intervene, the call is seamlessly handed over to the user. The user can then acknowledge the notification and respond directly.

[0747] Specific example

[0748] For example, consider a scenario where a customer calls to complain, saying, "The item I ordered hasn't arrived yet!"

[0749] The device sends the audio data to the server.

[0750] The server converts the audio data into text and analyzes it using an NLP engine.

[0751] The server detects keywords such as "order," "not received," and "dissatisfaction," and uses a generation AI to generate a response such as "We apologize, we will check on this immediately."

[0752] The emotion engine detects an angry tone and carefully adjusts its response accordingly.

[0753] The server sends the response to the customer via the terminal.

[0754] Based on sentiment data and keywords, the server sends a notification to the user stating, "We are handling a complaint; intervention is required."

[0755] The user reviews the notification and chooses to intervene. The system responds directly to the customer with, "We will investigate immediately, please wait."

[0756] Example of a prompt

[0757] "We have received an inquiry from a customer regarding a non-delivery of their product, and their message is filled with anger. Please prepare an appropriate response."

[0758] "The customer is dissatisfied with the delay in their order. Please generate a polite response to improve this situation."

[0759] In this way, the system of the present invention incorporates emotion recognition functionality to provide appropriate responses based on the user's emotions, and allows the user to intervene as needed. This enables improved efficiency and quality in telephone communication.

[0760] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0761] Step 1:

[0762] The device detects incoming calls and, once a call begins, sends audio data to the server in real time. Specifically, the device's phone function detects the start of a call, converts the audio picked up by the microphone into digital audio data, and sends it to the server via the internet.

[0763] Input: Incoming phone call

[0764] Output: Audio data transmitted in real time

[0765] Step 2:

[0766] The server receives audio data and converts it into text data in real time using a speech recognition engine (e.g., Google Cloud Speech-to-Text API). Specifically, the speech recognition engine analyzes the audio signal and converts the words used in the conversation into text format.

[0767] Input: Audio data transmitted in real time

[0768] Output: Text data

[0769] Step 3:

[0770] The server passes the converted text data to a natural language processing engine (e.g., Google Cloud Natural Language API) to analyze the conversation. Specifically, the natural language processing engine extracts meaning from the text data and identifies important keywords and topics.

[0771] Input: Text data converted by the speech recognition engine

[0772] Output: Analyzed keyword and topic information

[0773] Step 4:

[0774] The server uses a generative AI model (e.g., OpenAI's GPT-4) based on the analysis results to generate an appropriate response in text format. It utilizes prompts to adjust the generated response as needed.

[0775] Input: Analyzed keyword and topic information

[0776] Output: Generated text response

[0777] Example prompt:

[0778] "We have received an inquiry from a customer regarding a non-delivery of their product, and their message is filled with anger. Please prepare an appropriate response."

[0779] Step 5:

[0780] The server passes the generated text response to a text-to-speech engine (e.g., Microsoft Azure Text-to-Speech), which converts it into speech data in real time. Specifically, the text-to-speech engine converts the text into speech, creating natural and fluent speech output.

[0781] Input: Generated text response

[0782] Output: Generated audio data

[0783] Step 6:

[0784] The server passes the voice and text data to an emotion recognition engine (e.g., IBM Watson Tone Analyzer) to analyze the user's emotions. Specifically, the emotion recognition engine identifies emotions from the tone and word choice of the voice and then identifies areas that need adjustment based on that.

[0785] Input: Generated text response and audio data

[0786] Output: Emotion analysis results

[0787] Step 7:

[0788] The server adjusts the generated response based on the recognized emotion. For example, if anger is detected, the response will be made more careful and polite.

[0789] Input: Sentiment analysis results

[0790] Output: Adjusted text response and audio data

[0791] Step 8:

[0792] The server sends the generated audio data to the terminal, and the terminal sends it back to the person on the other end of the line. Specifically, the server sends audio data to the terminal in real time, and the terminal transmits that data to the person on the other end of the line through its audio output device (speaker).

[0793] Input: Adjusted audio data

[0794] Output: Transmitted audio data

[0795] Step 9:

[0796] Based on text data and sentiment analysis results, the server sends notifications to the user when specific keywords or important emotions are detected. These notifications may include alerts such as, "We are handling a complaint. Intervention is required."

[0797] Input: Specific keywords and sentiment analysis results

[0798] Output: User statement notification

[0799] Step 10:

[0800] The device displays a notification to the user and offers an option to intervene. If the user chooses to intervene, the call is seamlessly handed over to the user. Specifically, the user acknowledges the notification and responds directly to the call if necessary.

[0801] Input: User Statement Notification

[0802] Output: User intervention to handle calls

[0803] (Application Example 2)

[0804] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0805] Traditional telephone answering systems failed to consider customer emotions, and inappropriate responses easily led to customer dissatisfaction. Furthermore, automated answering systems could only handle a limited range of questions, requiring human intervention for complex inquiries and complaints, and this intervention was sometimes not timely. In addition, content delivery services require prompt and appropriate responses to inquiries to improve the user experience.

[0806] The specific processing performed by the specific 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 receiving voice data and converting it into text data in real time, means for analyzing the text data and generating a response using a generation AI model based on the analysis results, means for converting the generated response into voice data, means for analyzing the user's emotions from the text data and voice data using an emotion engine, means for transmitting the voice data of the generated response, and means for notifying the user when specific emotions and keywords are detected and providing an option for intervention. This provides an appropriate automated response that takes the user's emotions into consideration, and allows for human intervention at the appropriate time as needed.

[0807] "Audio data" refers to information that represents telephone calls or recorded audio in digital format.

[0808] "Text data" refers to character information generated by analyzing audio data and using speech recognition technology.

[0809] A "generative AI model" is a type of artificial intelligence that automatically generates appropriate responses to specific inputs (prompt sentences) based on natural language processing.

[0810] A "prompt statement" is an input statement used to generate a response for a generative AI model.

[0811] An "emotion engine" is an algorithm or system that analyzes voice and text data to recognize the user's emotions (such as joy, anger, or sadness) contained within it.

[0812] "Analysis results" refer to conclusions or information obtained through analysis of text data using natural language processing or sentiment engines.

[0813] "Response" refers to the reply or action generated by a generative AI model based on its analysis results.

[0814] A "notification" is an alert or message sent from a system to a user under specific conditions (for example, detection of the user's emotions or keywords).

[0815] "Intervention" refers to the act of a user directly participating in a call or response when the system cannot handle the situation through automated responses or under specific conditions.

[0816] "Means" refer to equipment, algorithms, systems, methods, etc., used to achieve a specific purpose.

[0817] The present invention relates to a system that automates telephone answering and recognizes user emotions to adjust responses. The system comprises means for receiving voice data and converting it into text data in real time, means for analyzing the text data and generating responses using a generative AI model, means for converting the generated responses into voice data, means for analyzing the user's emotions using an emotion engine, means for transmitting the generated voice data, and means for notifying the user when specific emotions or keywords are detected and providing options for intervention.

[0818] System Overview

[0819] Receiving and converting audio data:

[0820] The device detects an incoming call and sends audio data to the server in real time when the call begins. The server uses a speech recognition engine (e.g., Google Speech Recognition) to convert the audio data into text data.

[0821] Text data analysis and response generation:

[0822] The server passes text data to an NLP engine, which analyzes the content and intent of the conversation. Based on the analysis results, it generates a response using a generative AI model (e.g., the OpenAI GPT model). At this time, prompts are used to guide the model in generating the response.

[0823] Emotion recognition by an emotion engine:

[0824] Text and audio data are passed to an emotion engine (e.g., Hugging Face's sentiment-analysis pipeline) to analyze the user's emotions. Based on the recognized emotions, the server adjusts its response.

[0825] Sending responses and notifications:

[0826] The generated audio data is sent to the terminal and then forwarded from the terminal to the customer. Simultaneously, the server detects specific emotions and keywords, and if important call content or emotions are detected, it notifies the user. The user reviews the notification, and if they choose to intervene, the call is seamlessly handed over to them.

[0827] Specific example

[0828] For example, if a customer says, "The video won't play," the system will process it as follows:

[0829] The terminal sends audio data to the server, and the server converts the audio data into text.

[0830] The system analyzes text data and uses a generative AI model to generate an appropriate response ("We apologize, we will investigate").

[0831] The emotion engine detects user dissatisfaction, and the server adjusts its response to be more attentive.

[0832] The server converts the response into audio data and sends it to the customer from the terminal.

[0833] When emotions and keywords are detected, a notification is sent to the user, and they are given the option to intervene.

[0834] Example of a prompt:

[0835] A user said: "The video won't play." How do you respond?

[0836] The system described above enables prompt and appropriate responses to customer inquiries and complaints. Furthermore, by providing responses that take user emotions into account, improved customer satisfaction can be expected.

[0837] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0838] Step 1:

[0839] The terminal detects an incoming call and receives audio data in real time when the call begins. It then sends the received audio data to the server. Specifically, the terminal uses a microphone to capture audio and transfers the audio data to the server in digital format. In this step, the input is the audio data, and the output is the audio data sent to the server.

[0840] Step 2:

[0841] The server converts the received audio data into text data in real time using a speech recognition engine (e.g., Google Speech Recognition). The server converts the audio to text using digital signal processing and further generates text data for the next processing step. In this step, the input is audio data and the output is text data.

[0842] Step 3:

[0843] The server passes text data to a natural language processing engine for analysis. During the analysis process, it extracts specific intentions and themes and generates data to provide to the AI ​​model that generates appropriate prompt sentences. For example, the keyword "cannot be played" may be detected from the text data. In this step, the input is text data, and the output is the analysis results and prompt sentences for the AI ​​model.

[0844] Step 4:

[0845] The server inputs a prompt message into the generative AI model, which then generates an appropriate response. The generative AI model generates a text-based response based on the provided prompt message. This text-based response contains the solution to the user's question or problem. In this step, the input is the prompt message, and the output is the text data of the generated response.

[0846] Step 5:

[0847] The server passes the generated response text data to a speech synthesis engine (e.g., pyttsx3) to convert it into speech data. The speech synthesis engine converts the text into natural-sounding speech, making it ready to send the response to the customer. In this step, the input is the response text data, and the output is the response speech data.

[0848] Step 6:

[0849] The server passes text and audio data to the emotion engine (e.g., Hugging Face's sentiment-analysis pipeline) to analyze the user's emotions. This analysis identifies the user's positive, negative, or neutral emotions and may reflect them in the response. In this step, the input is text and audio data, and the output is the emotion analysis result.

[0850] Step 7:

[0851] The server further refines the generated response based on the sentiment analysis results, providing it in a way that is appropriate to the user's emotions. For example, if the user is angry, the response is adjusted to be more polite. In this step, the input is the sentiment analysis results and the generated response text data, and the output is the adjusted response text data.

[0852] Step 8:

[0853] The server sends the finalized response audio data to the terminal, which then sends it back to the caller. The terminal plays the audio back to the customer in real time, conveying the content of the response. In this step, the input is the adjusted response audio data, and the output is the response to the customer.

[0854] Step 9:

[0855] If specific emotions and keywords are detected, the server sends a notification to the user informing them that intervention is needed. The user reviews the notification and chooses whether to intervene in the call if necessary. In this step, the inputs are the emotion analysis results and keyword detection results, and the outputs are the notification to the user and the choice of intervention.

[0856] 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 user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0857] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0858] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart glasses 214.

[0859] [Third Embodiment]

[0860] Figure 5 shows an example of the configuration of the data processing system 310 according to the third embodiment.

[0861] As shown in Figure 5, the data processing system 310 includes a data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0862] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0863] The headset terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and display 343 are also connected to the bus 52.

[0864] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0865] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0866] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0867] Figure 6 shows an example of the main functions of the data processing device 12 and the headset terminal 314. As shown in Figure 6, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0868] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[0869] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0870] In the headset terminal 314, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0871] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the headset terminal 314 will be referred to as the "terminal".

[0872] The system according to the present invention receives voice data, converts it to text data in real time, analyzes it, generates an appropriate response, and transmits it as voice data. It also detects specific keywords and notifies the user, enabling seamless switching so that the user can intervene as needed.

[0873] System-wide configuration

[0874] The system consists of the following elements:

[0875] 1. Terminal

[0876] 2. Server

[0877] 3. User

[0878] Receiving and converting audio data

[0879] The device detects an incoming call and, as soon as the call begins, sends the audio data to the server in real time.

[0880] The server passes the received audio data to the speech recognition engine, which converts it into text data in real time.

[0881] Text data analysis and response generation

[0882] The server uses a natural language processing (NLP) engine to analyze text data and understand its content.

[0883] Based on the analysis results, the server automatically generates an appropriate response using a generation AI. This response is then generated again in text format.

[0884] The server passes the text-based response to the speech synthesis engine, which converts it into speech data.

[0885] Sending a response and notification

[0886] The server sends the generated voice data to the terminal, which then relays it to the person on the other end of the line.

[0887] The server continuously monitors the call content and notifies the user via the terminal if it detects specific keywords.

[0888] Seamless user intervention

[0889] Users can check notifications and, if important call content is detected, intervene in the call themselves.

[0890] The device accepts user intervention, allowing the user to speak directly with the person they are calling.

[0891] Specific example

[0892] 1. Handling sales calls

[0893] A salesman makes a phone call and says, "Excuse me for bothering you while you're busy, this is a salesperson from AA Insurance."

[0894] The device transmits voice data to the server in real time.

[0895] The server converts the voice data into text and recognizes it as "This is a salesperson from AA Insurance."

[0896] The server uses a generation AI to generate the response, "I'm not interested, thank you," converts it into audio data, and sends it to the terminal.

[0897] The device responds to the salesman by playing this audio data.

[0898] 2. Important notifications and user intervention

[0899] The salesman continued, "Would you mind if I sent you some materials?"

[0900] The server detects specific keywords such as "send documents" and notifies the user via the terminal.

[0901] The user checks the notification and taps the screen to intervene in the call, directly responding, "I don't need to send any documents."

[0902] Thus, the present invention is a system that automates telephone answering while allowing user intervention in important situations, thereby achieving efficient and flexible responses.

[0903] Next, we can describe the details of the processing steps.

[0904] The following describes the processing flow.

[0905] Step 1:

[0906] The device detects an incoming call and collects the call audio as soon as the call begins. The collected audio data is transmitted to the server in real time.

[0907] Step 2:

[0908] The server passes the received audio data to the speech recognition engine, which converts the audio data into text data. This conversion is performed in real time and processed sequentially.

[0909] Step 3:

[0910] The server passes the text data to a natural language processing (NLP) engine for content analysis. The analysis identifies the purpose and topic of the call.

[0911] Step 4:

[0912] The server then uses the generation AI to create a response based on the analysis results. This response is generated in text format and designed to be an appropriate reply to the conversation.

[0913] Step 5:

[0914] The server passes the generated text response to the speech synthesis engine, which converts it into speech data. The converted speech data is then generated again in real time.

[0915] Step 6:

[0916] The server sends the audio data to the terminal. The terminal decodes the received audio data and plays it back to the caller at the appropriate time.

[0917] Step 7:

[0918] The server continues to monitor the call content to check for the inclusion of specific keywords or phrases. For example, it may detect specific keywords such as "important matter" or "sending documents."

[0919] Step 8:

[0920] When a specific keyword is detected, the server sends a notification to the terminal, informing the user that intervention is necessary.

[0921] Step 9:

[0922] The device displays a notification to the user, alerting them with audio and visuals. It then prompts the user to choose whether or not to take action.

[0923] Step 10:

[0924] The user checks the notification and, if important call content is detected, intervenes in the call at their own discretion. The device accepts the user's intervention and allows the user to speak directly with the other party.

[0925] Step 11:

[0926] Once the call ends, the device sends the audio recording and text data to the server. The server stores this data in a database for later reference.

[0927] In this way, the system seamlessly integrates real-time voice responses with user intervention, enabling effective telephone handling.

[0928] (Example 1)

[0929] Next, we will describe Example 1. 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."

[0930] Conventional voice response systems were unable to adequately generate responses or detect keywords in real time, making it difficult to respond immediately to important call content. Furthermore, the inability to seamlessly integrate automated responses with user intervention made flexible and efficient call handling challenging.

[0931] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0932] In this invention, the server includes means for receiving voice data and converting it into text data in real time; means for analyzing the text data and generating an appropriate response using a generative AI model; means for converting the generated text data of the response into voice data; means for transmitting the generated voice data of the response; means for detecting specific keywords and notifying the user; and means for allowing the user to acknowledge the notification and intervene in the call. This enables real-time text conversion and response generation, keyword detection and user notification, and seamless user intervention.

[0933] "Audio data" refers to audio signals obtained through phone calls, recordings, etc., represented in digital format.

[0934] "Real-time" refers to data processing and responses occurring immediately, with results reflected without delay.

[0935] "Text data" refers to audio data converted into written information, expressed in a format that is understandable to humans.

[0936] A "generative AI model" refers to a technology that uses artificial intelligence algorithms to generate appropriate responses or data based on given input.

[0937] A "prompt" is a series of text inputs to a generative AI model, which the AI ​​uses to generate a response.

[0938] "Analysis" is a means of processing text data to understand its content and structure, and a process for deriving appropriate responses or actions.

[0939] A "keyword" is an important word or phrase contained within text data, and its detection triggers a specific action or response.

[0940] A "user" is a person who uses the system and has the authority to intervene in the content of a call.

[0941] A "notification" is an informational alert sent from the system to the user, intended to draw the user's attention when a specific keyword is detected.

[0942] "Intervening in a call" refers to the action of a user switching from an automated system response to a manual response and starting a direct conversation with the other party.

[0943] A "speech recognition engine" is software or hardware that analyzes speech data and converts it into text data.

[0944] A "speech synthesis engine" is software or hardware that analyzes text data and converts it into speech data.

[0945] "Communication methods" refer to technologies used to send and receive voice and text data between a server and a terminal, and include the Internet Protocol and other communication protocols.

[0946] A specific embodiment of the system according to the present invention is shown below. This system receives voice data, converts it into text data in real time, performs subsequent analysis and response generation, and enables notifications and call interventions to the user.

[0947] Receiving and converting audio data

[0948] The terminal detects an incoming call and captures audio data as soon as the call begins. The call audio is transmitted to the server in real time. WebSocket and HTTP streaming technologies are used as secure communication methods for transmitting the audio data.

[0949] The server converts the received audio data into text data using the Google Cloud Speech-to-Text API or Amazon Transcribe. The converted text data is extracted as a string that accurately represents the spoken content.

[0950] Text data analysis and response generation

[0951] The server uses Google Cloud Natural Language API and Microsoft Azure Text Analytics to analyze text data and understand its content. The information extracted through analysis includes grammatical structure, meaning, and important keywords.

[0952] Based on the analysis results, the server automatically generates an appropriate response using a generative AI model (e.g., GPT-3). The generative AI model used here has advanced natural language generation capabilities and generates the optimal response from the given prompt sentence.

[0953] The generated response is then converted back into text format and then into speech data using the Google Text-to-Speech API or Amazon Polly.

[0954] Sending a response and notification

[0955] The server sends the generated audio data to the terminal, which then relays that audio data to the person on the other end of the call. This enables real-time responses.

[0956] A server that continuously monitors call content will notify the user via the terminal if it detects a specific keyword, such as the phrase "send documents." The notification system can display the notification on the terminal screen or alert the user with vibration or sound.

[0957] Seamless user intervention

[0958] Upon receiving a notification, the user can review the notification displayed on their device screen and decide whether or not to intervene in the call. If the user chooses to intervene, tapping the notification on their device will switch the system from automatic response to manual response, allowing the user to speak directly with the caller.

[0959] Specific example

[0960] 1. Handling sales calls

[0961] A salesman makes a phone call and says, "Excuse me for bothering you while you're busy, this is a salesperson from AA Insurance."

[0962] The device transmits voice data to the server in real time.

[0963] The server converts the voice data into text and recognizes it as "This is a salesperson from AA Insurance."

[0964] The server uses a generation AI to generate the response, "I'm not interested, thank you," converts it into audio data, and sends it to the terminal.

[0965] The device responds to the salesman by playing this audio data.

[0966] 2. Important notifications and user intervention

[0967] The salesman continued, "Would you mind if I sent you some materials?"

[0968] The server detects specific keywords such as "send documents" and notifies the user via the terminal.

[0969] The user checks the notification, taps the screen to intervene in the call, and directly responds, "I don't need to send any documents."

[0970] Example of a prompt

[0971] "If a salesperson is soliciting insurance and includes the phrase 'send me some materials,' the user should be notified. The response should generate 'I do not need any materials.'"

[0972] This system enables efficient and flexible call handling by transcribing audio in real time, generating appropriate responses, and allowing user intervention at key points.

[0973] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0974] Step 1:

[0975] The device detects an incoming call and captures audio data as soon as the call begins. The audio data is sent to the server in real time. Specifically, it receives the phone's audio in real time as input and sends that audio data to the server via the network as output. In practice, the smartphone outputs a log message saying "Audio capture started" and immediately begins sending the audio data.

[0976] Step 2:

[0977] The server passes the received audio data to a speech recognition engine (e.g., Google Cloud Speech-to-Text API) and converts it into text data in real time. It takes audio data as input and generates text data as output. Specifically, the server outputs a log message saying "Audio data converted to text."

[0978] Step 3:

[0979] The server uses a natural language processing (NLP) engine (e.g., Google Cloud Natural Language API) to analyze text data. It receives text data as input and generates analysis results as output. Analysis includes entity extraction and sentiment analysis. Specifically, the server logs "Text analysis complete: Intent is sales."

[0980] Step 4:

[0981] The server generates an appropriate response based on the analysis results using a generative AI model (e.g., GPT-3). It receives the analysis results and a prompt as input and generates the appropriate response text as output. Specifically, the server logs "Response generation complete." An example of a prompt is: "If a salesman is soliciting insurance and includes the phrase 'send materials,' the user should be notified. The response should generate 'No materials are needed.'"

[0982] Step 5:

[0983] The server passes the generated response text to a text-to-speech engine (e.g., Google Text-to-Speech API) to convert it into speech data. It takes response text as input and generates speech data as output. Specifically, the server logs "Text-to-speech complete".

[0984] Step 6:

[0985] The server sends the generated audio data to the terminal, and the terminal transmits that audio data to the person it is talking to. It receives audio data as input and plays it back to the person it is talking to as output. Specifically, the terminal log displays "Playing response audio".

[0986] Step 7:

[0987] The server continuously monitors the call content and, if it detects a specific keyword (e.g., "send documents"), it notifies the user via the terminal. It receives real-time text data as input and generates a notification as output. Specifically, the server log records "Keyword detected: Send documents" and sends a notification to the terminal.

[0988] Step 8:

[0989] The user can check notifications and, if important call content is detected, intervene in the call themselves. The device receives notifications as input and intervenes in the call as output. Specifically, the user taps the notification on their device and directly responds with "No documents are needed." The device then logs "User intervened in the call."

[0990] (Application Example 1)

[0991] Next, we will explain Application Example 1. In the following explanation, 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."

[0992] Traditional content distribution services have faced challenges in providing personalized recommendations based on user preferences and interests, making it difficult to respond quickly and accurately to the information and entertainment users seek. Furthermore, the lack of systems that use AI-generated models to respond to requests has made it difficult to obtain the information users need in real time.

[0993] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0994] In this invention, the server includes means for receiving voice data and converting it into text data in real time; means for analyzing the text data, generating a response using a generative AI model based on the analysis results, and converting the response into voice data; means for transmitting the generated voice data of the response; means for obtaining a request from the user in voice data and sending a prompt sentence to the generative AI model based on the request; and means for suggesting appropriate content to the user based on the generated prompt. This makes it possible to recommend the most suitable content to the user in real time.

[0995] "Audio data" refers to data obtained by converting audio into a digital format.

[0996] "Real-time" refers to processing data instantly and without delay.

[0997] "Text data" refers to digital data obtained by converting audio data into written text.

[0998] A "generative AI model" refers to an artificial intelligence model that generates appropriate responses based on user input data.

[0999] A "prompt" is a portion of the text data input to a generating AI model, and it functions as an instruction or question for the model.

[1000] "Analysis" refers to the process of understanding text data and extracting its meaning.

[1001] "Content" is a general term for information and entertainment that users consume, such as movies, music, and news.

[1002] "Suggestions" refer to the act of recommending appropriate content based on the user's preferences and requests.

[1003] This invention relates to a system that converts audio data into text data in real time, analyzes the text data to generate a response using a generative AI model, and transmits that response back as audio data. Furthermore, it also includes a function to generate prompt sentences for the generative AI model based on user requests and suggest appropriate content.

[1004] System Configuration

[1005] The system consists of the following elements:

[1006] 1. Terminal

[1007] An input device for receiving audio data (e.g., a microphone)

[1008] An output device for playing audio data (e.g., a speaker)

[1009] 2. Server

[1010] Speech recognition engine

[1011] Natural Language Processing (NLP) Engine

[1012] Generative AI models (e.g., GPT-3)

[1013] Speech synthesis engine

[1014] 3. User

[1015] Person making a request by voice

[1016] Receiving and converting audio data

[1017] The terminal receives the user's voice data and sends it to the server in real time. The server uses a speech recognition engine to convert the voice data into text data.

[1018] Text data analysis and response generation

[1019] The server analyzes the converted text data using a natural language processing (NLP) engine to understand its content. Based on the analysis results, it uses a generative AI model to generate an appropriate response, which is then generated again in text format. The generated text is then converted into speech data by a speech synthesis engine.

[1020] Sending a response and suggesting content

[1021] The generated audio data is sent to the device and transmitted to the user through the device's speaker. Furthermore, a request is obtained from the user via the audio data, and a prompt is sent to the generating AI model based on that request. Based on the generated prompt, appropriate content is suggested to the user.

[1022] Specific example

[1023] For example, if a user requests "Tell me your recommendations for the latest movies," the following prompt message will be generated.

[1024] Examples of prompts to input into a generative AI model:

[1025] User request: Please recommend some of the latest movies.

[1026] Appropriate response:

[1027] Based on this prompt, the server sends information to the AI ​​model that generates the appropriate response text. For example, a response such as "As a recent movie recommendation, we suggest the action film XXX" is generated, converted into audio data, and delivered to the user.

[1028] This allows users to receive real-time responses to their requests and to have content that interests them suggested quickly and accurately.

[1029] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[1030] Step 1:

[1031] The device receives the user's voice data.

[1032] Input: User's voice data

[1033] Operation: The microphone captures the user's voice and converts the audio data into a digital format.

[1034] Output: Digital audio data

[1035] Step 2:

[1036] The terminal sends the received digital audio data to the server.

[1037] Input: Digital audio data

[1038] Operation: Transmits audio data to the server in real time via the network interface.

[1039] Output: Audio data sent to the server

[1040] Step 3:

[1041] The server uses a speech recognition engine to convert the audio data into text data.

[1042] Input: Audio data

[1043] Operation: Analyzes audio data using a speech recognition engine (e.g., Google Speech Recognition) and converts it to text.

[1044] Output: Text data

[1045] Step 4:

[1046] The server analyzes the text data using a natural language processing (NLP) engine and understands its meaning.

[1047] Input: Text data

[1048] Operation: Uses a natural language processing (NLP) engine to extract meaning from text data.

[1049] Output: Analysis results (user request details)

[1050] Step 5:

[1051] Based on the analysis results, the server generates a response using an AI model.

[1052] Input: Analysis results

[1053] Operation: A prompt is sent to a generative AI model (e.g., GPT-3), and a response text is generated based on it. An example prompt is "User request: Tell me your recommendations for the latest movies. Appropriate response:"

[1054] Output: Response text data

[1055] Step 6:

[1056] The server converts the generated response text into speech data using a speech synthesis engine.

[1057] Input: Response text data

[1058] Operation: Uses a speech synthesis engine (e.g., pyttsx3) to convert response text into speech data.

[1059] Output: Response voice data

[1060] Step 7:

[1061] The server sends the generated response voice data to the terminal.

[1062] Input: Response voice data

[1063] Operation: Transmits voice data to the terminal in real time via the network interface.

[1064] Output: Response voice data sent to the terminal

[1065] Step 8:

[1066] The device plays back the received voice response data and conveys it to the user.

[1067] Input: Response voice data

[1068] Operation: Plays audio data through the speaker.

[1069] Output: The user receives a voice response.

[1070] Step 9:

[1071] If a new request comes from the user, the same process is repeated.

[1072] Input: New voice request

[1073] Action: Repeat the process from step 1.

[1074] Output: New response to the user

[1075] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[1076] The system related to this invention automates telephone answering and allows for seamless user intervention as needed. Furthermore, this system incorporates an emotion engine that recognizes the user's emotions and adjusts its response accordingly.

[1077] System-wide configuration

[1078] The system consists of the following elements:

[1079] 1. Terminal

[1080] 2. Server

[1081] 3. User

[1082] 4. Emotional Engine

[1083] Receiving and converting audio data

[1084] The device detects an incoming call and sends audio data to the server in real time when the call begins. The audio data is the content of the call and is sent to the server sequentially.

[1085] Text data analysis and response generation

[1086] The server receives the audio data and converts it into text data in real time using a speech recognition engine.

[1087] The server passes the text data to the NLP engine, which analyzes the content and intent of the conversation. This clarifies the specific topic and purpose of the call.

[1088] Based on the analysis results, the server uses a generative AI to generate an appropriate response in text format. This process builds an automated response system for calls.

[1089] The server passes the generated text response to the speech synthesis engine, which converts it into speech data in real time.

[1090] Emotion recognition by an emotion engine

[1091] The server passes text and audio data to the emotion engine, which analyzes the user's emotions. For example, emotions can be inferred from the tone of voice and word choice.

[1092] Based on the emotions recognized by the emotion engine, the server adjusts its response. If the user is angry, a more careful response will be generated.

[1093] The server detects specific emotions and keywords and sends a notification to the user based on them. In this case, it is determined that the user needs to intervene.

[1094] Sending a response and notification

[1095] The server sends the generated audio data to the terminal, which then sends it back to the person on the other end of the call. Simultaneously, if important conversation content or emotions are detected, the user is notified.

[1096] The device displays a notification to the user and offers them an option to intervene. If the user chooses to intervene, the call is seamlessly handed over to the user.

[1097] Specific example

[1098] 1. Examples of handling customer complaints

[1099] A customer calls to complain: "The item I ordered hasn't arrived yet!"

[1100] The device sends the audio data to the server.

[1101] The server converts the audio data into text and analyzes it using an NLP engine.

[1102] The server detects keywords such as "order," "not received," and "dissatisfaction," and uses a generation AI to generate a response such as "We apologize, we will check on this immediately."

[1103] The emotion engine detects an angry tone and carefully adjusts its response accordingly.

[1104] The server sends the response to the customer via the terminal.

[1105] The server sends a notification to the user based on sentiment data and keywords. It displays a message saying, "We are handling a complaint and intervention is required."

[1106] The user reviews the notification and selects intervention. The system responds directly to the customer with, "We will investigate immediately, please wait."

[1107] In this way, the system of the present invention incorporates emotion recognition functionality to provide appropriate responses based on the user's emotions, and allows the user to intervene as needed. This enables improved efficiency and quality in telephone communication.

[1108] The following describes the processing flow.

[1109] Step 1:

[1110] The device detects incoming calls and collects call audio as soon as the call begins. The collected audio data is sent to the server in real time.

[1111] Step 2:

[1112] The server passes the received audio data to the speech recognition engine, which converts the audio data into text data. This conversion is performed in real time and processed sequentially.

[1113] Step 3:

[1114] The server then passes the text data to a natural language processing (NLP) engine for analysis. The analysis results identify the purpose and topic of the call.

[1115] Step 4:

[1116] The server passes the analyzed text data to the emotion engine, which analyzes the user's emotions. The emotion engine identifies the user's emotions based on their tone of voice and word choice.

[1117] Step 5:

[1118] The server integrates the results from the emotion engine and the NLP engine's analysis, and has the generative AI create an appropriate response. This response is generated in text format and is designed to be an appropriate reply in a conversation.

[1119] Step 6:

[1120] The server passes the generated text response to the speech synthesis engine, which converts it into speech data. The generated speech data is then generated again in real time.

[1121] Step 7:

[1122] The server sends the generated voice data to the terminal, which then sends it to the caller. This is where the response is conveyed to the other party.

[1123] Step 8:

[1124] The server continues to monitor the call content and sends notifications to the user if specific keywords or emotions are detected.

[1125] Step 9:

[1126] The device displays notifications to the user, alerting them with audio and visuals. These notifications may include messages such as, "A customer is expressing dissatisfaction. Intervention is needed."

[1127] Step 10:

[1128] The user reviews the notification and, if important call content or emotions are detected, intervenes in the call at their own discretion. The device accepts the user's intervention.

[1129] Step 11:

[1130] The user chooses to intervene and begins a direct conversation with the person on the other end of the line. This allows for a seamless integration of human judgment into responses, even when emotional responses are required.

[1131] Step 12:

[1132] Once the call ends, the device sends the audio recording and text data to the server. The server stores this data in a database for later reference.

[1133] By combining this with an emotion engine, it becomes possible to automate telephone answering while still providing flexible responses based on the user's emotions.

[1134] (Example 2)

[1135] Next, we will describe Example 2. 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."

[1136] Traditional telephone answering systems often relied on manual processes for analyzing call content and generating appropriate responses, resulting in challenges in response quality and efficiency. Furthermore, they struggled to respond in a way that considered user emotions, hindering responses that required quick and appropriate action, particularly in situations like complaint handling. Additionally, delays in detecting specific key keywords or emotions could lead to delayed user intervention and inadequate responses.

[1137] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[1138] In this invention, the server includes means for detecting an incoming call and transmitting voice data in real time when the call begins; means for receiving the voice data and converting it into text data in real time using recognition means; means for passing the text data to analysis means and creating a response based on the analysis results using generation means; means for passing the text data of the generated response to synthesis means and converting it into voice data; and means for transmitting the generated voice data. This enables real-time analysis of call content and automatic generation of appropriate responses.

[1139] "Means for detecting incoming phone calls" refers to devices or software that identify call requests received via telephone lines or internet communication.

[1140] "Means of transmitting voice data in real time" refers to devices or software that have the function of instantly processing voice information spoken during a call and transmitting it to other devices or systems via a network.

[1141] "Means for receiving audio data and converting it into text data in real time using recognition means" refers to devices or software that have the function of instantly converting received audio information into text data using a speech recognition engine.

[1142] "Means for passing text data to an analysis means and creating a response based on the analysis results using a generation means" refers to devices or software that send text data to a natural language processing engine and create an appropriate response sentence using a generation AI or similar based on the analysis results.

[1143] "Means for passing the generated response text data to a synthesis means and converting it into audio data" refers to devices or software that have the function of sending a text-based response to a speech synthesis engine and converting it into a format that can be played back as audio information.

[1144] "Means for transmitting generated audio data" refers to devices or software that have the function of transmitting audio data generated by a speech synthesis engine to a designated terminal or receiving device.

[1145] "Means for passing text data and audio data to emotion recognition means and analyzing the user's emotions" refers to devices or software that have the function of determining emotions from text and audio using an analysis engine and identifying the speaker's emotional state.

[1146] "Means of adjusting response content based on emotions" refers to devices or software that have the function of modifying responses to appropriate tone and wording based on emotional information identified by an analysis engine.

[1147] "Means of notifying a user when specific keywords are detected, allowing the user to intervene in a call" refers to devices or software that have the function of notifying an administrator or operator when pre-set important keywords or emotional states are detected, providing them with an opportunity to intervene in the situation.

[1148] The system related to this invention automates telephone answering, allowing user intervention as needed. Furthermore, the system incorporates an emotion engine that recognizes the user's emotions and adjusts its response accordingly. This system includes the following elements:

[1149] 1. Terminal

[1150] 2. Server

[1151] 3. User

[1152] 4. Emotional Engine

[1153] Receiving and converting audio data

[1154] The device detects an incoming call and, once the call begins, sends audio data to the server in real time. This audio data, containing the conversation content, is transmitted to the server continuously. Specifically, smartphones and landline phones fulfill this role.

[1155] Text data analysis and response generation

[1156] The server receives the audio data and converts it into text data in real time using a speech recognition engine (e.g., Google Cloud Speech-to-Text API). The converted text data is then passed to a natural language processing engine (e.g., Google Cloud Natural Language API) to analyze the content and intent of the conversation.

[1157] Next, based on the analysis results, a generative AI (e.g., OpenAI's GPT-4) is used to generate an appropriate response in text format. The generated text response is then passed to a speech synthesis engine (e.g., Microsoft Azure Text-to-Speech) and converted into speech data in real time.

[1158] Emotion recognition and response adjustment by an emotion engine

[1159] The server passes text and audio data to an emotion engine (e.g., IBM Watson Tone Analyzer) to analyze the user's emotions. Based on the analysis, the response is adjusted. For example, if the user is angry, the response will be more careful and polite.

[1160] Sending a response and notification

[1161] The server sends the generated audio data to the terminal, which then sends it back to the caller. Simultaneously, if important call content or emotions are detected, the user is notified. The notification might say something like, "We are handling a complaint. Intervention is required."

[1162] User intervention

[1163] The device displays a notification to the user and offers an option to intervene. If the user chooses to intervene, the call is seamlessly handed over to the user. The user can then acknowledge the notification and respond directly.

[1164] Specific example

[1165] For example, consider a scenario where a customer calls to complain, saying, "The item I ordered hasn't arrived yet!"

[1166] The device sends the audio data to the server.

[1167] The server converts the audio data into text and analyzes it using an NLP engine.

[1168] The server detects keywords such as "order," "not received," and "dissatisfaction," and uses a generation AI to generate a response such as "We apologize, we will check on this immediately."

[1169] The emotion engine detects an angry tone and carefully adjusts its response accordingly.

[1170] The server sends the response to the customer via the terminal.

[1171] Based on sentiment data and keywords, the server sends a notification to the user stating, "We are handling a complaint; intervention is required."

[1172] The user reviews the notification and chooses to intervene. The system responds directly to the customer with, "We will investigate immediately, please wait."

[1173] Example of a prompt

[1174] "We have received an inquiry from a customer regarding a non-delivery of their product, and their message is filled with anger. Please prepare an appropriate response."

[1175] "The customer is dissatisfied with the delay in their order. Please generate a polite response to improve this situation."

[1176] In this way, the system of the present invention incorporates emotion recognition functionality to provide appropriate responses based on the user's emotions, and allows the user to intervene as needed. This enables improved efficiency and quality in telephone communication.

[1177] The flow of the specific processing in Example 2 will be explained using Figure 13.

[1178] Step 1:

[1179] The device detects incoming calls and, once a call begins, sends audio data to the server in real time. Specifically, the device's phone function detects the start of a call, converts the audio picked up by the microphone into digital audio data, and sends it to the server via the internet.

[1180] Input: Incoming phone call

[1181] Output: Audio data transmitted in real time

[1182] Step 2:

[1183] The server receives audio data and converts it into text data in real time using a speech recognition engine (e.g., Google Cloud Speech-to-Text API). Specifically, the speech recognition engine analyzes the audio signal and converts the words used in the conversation into text format.

[1184] Input: Audio data transmitted in real time

[1185] Output: Text data

[1186] Step 3:

[1187] The server passes the converted text data to a natural language processing engine (e.g., Google Cloud Natural Language API) to analyze the conversation. Specifically, the natural language processing engine extracts meaning from the text data and identifies important keywords and topics.

[1188] Input: Text data converted by the speech recognition engine

[1189] Output: Analyzed keyword and topic information

[1190] Step 4:

[1191] The server uses a generative AI model (e.g., OpenAI's GPT-4) based on the analysis results to generate an appropriate response in text format. It utilizes prompts to adjust the generated response as needed.

[1192] Input: Analyzed keyword and topic information

[1193] Output: Generated text response

[1194] Example prompt:

[1195] "We have received an inquiry from a customer regarding a non-delivery of their product, and their message is filled with anger. Please prepare an appropriate response."

[1196] Step 5:

[1197] The server passes the generated text response to a text-to-speech engine (e.g., Microsoft Azure Text-to-Speech), which converts it into speech data in real time. Specifically, the text-to-speech engine converts the text into speech, creating natural and fluent speech output.

[1198] Input: Generated text response

[1199] Output: Generated audio data

[1200] Step 6:

[1201] The server passes the voice and text data to an emotion recognition engine (e.g., IBM Watson Tone Analyzer) to analyze the user's emotions. Specifically, the emotion recognition engine identifies emotions from the tone and word choice of the voice and then identifies areas that need adjustment based on that.

[1202] Input: Generated text response and audio data

[1203] Output: Emotion analysis results

[1204] Step 7:

[1205] The server adjusts the generated response based on the recognized emotion. For example, if anger is detected, the response will be made more careful and polite.

[1206] Input: Sentiment analysis results

[1207] Output: Adjusted text response and audio data

[1208] Step 8:

[1209] The server sends the generated audio data to the terminal, and the terminal sends it back to the person on the other end of the line. Specifically, the server sends audio data to the terminal in real time, and the terminal transmits that data to the person on the other end of the line through its audio output device (speaker).

[1210] Input: Adjusted audio data

[1211] Output: Transmitted audio data

[1212] Step 9:

[1213] Based on text data and sentiment analysis results, the server sends notifications to the user when specific keywords or important emotions are detected. These notifications may include alerts such as, "We are handling a complaint. Intervention is required."

[1214] Input: Specific keywords and sentiment analysis results

[1215] Output: User statement notification

[1216] Step 10:

[1217] The device displays a notification to the user and offers an option to intervene. If the user chooses to intervene, the call is seamlessly handed over to the user. Specifically, the user acknowledges the notification and responds directly to the call if necessary.

[1218] Input: User Statement Notification

[1219] Output: User intervention to handle calls

[1220] (Application Example 2)

[1221] Next, we will explain application example 2. In the following explanation, 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."

[1222] Traditional telephone answering systems failed to consider customer emotions, and inappropriate responses easily led to customer dissatisfaction. Furthermore, automated answering systems could only handle a limited range of questions, requiring human intervention for complex inquiries and complaints, and this intervention was sometimes not timely. In addition, content delivery services require prompt and appropriate responses to inquiries to improve the user experience.

[1223] The specific processing performed by the specific 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 receiving voice data and converting it into text data in real time, means for analyzing the text data and generating a response using a generation AI model based on the analysis results, means for converting the generated response into voice data, means for analyzing the user's emotions from the text data and voice data using an emotion engine, means for transmitting the voice data of the generated response, and means for notifying the user when specific emotions and keywords are detected and providing an option for intervention. This provides an appropriate automated response that takes the user's emotions into consideration, and allows for human intervention at the appropriate time as needed.

[1224] "Audio data" refers to information that represents telephone calls or recorded audio in digital format.

[1225] "Text data" refers to character information generated by analyzing audio data and using speech recognition technology.

[1226] A "generative AI model" is a type of artificial intelligence that automatically generates appropriate responses to specific inputs (prompt sentences) based on natural language processing.

[1227] A "prompt statement" is an input statement used to generate a response for a generative AI model.

[1228] An "emotion engine" is an algorithm or system that analyzes voice and text data to recognize the user's emotions (such as joy, anger, or sadness) contained within it.

[1229] "Analysis results" refer to conclusions or information obtained through analysis of text data using natural language processing or sentiment engines.

[1230] "Response" refers to the reply or action generated by a generative AI model based on its analysis results.

[1231] A "notification" is an alert or message sent from a system to a user under specific conditions (for example, detection of the user's emotions or keywords).

[1232] "Intervention" refers to the act of a user directly participating in a call or response when the system cannot handle the situation through automated responses or under specific conditions.

[1233] "Means" refer to equipment, algorithms, systems, methods, etc., used to achieve a specific purpose.

[1234] The present invention relates to a system that automates telephone answering and recognizes user emotions to adjust responses. The system comprises means for receiving voice data and converting it into text data in real time, means for analyzing the text data and generating responses using a generative AI model, means for converting the generated responses into voice data, means for analyzing the user's emotions using an emotion engine, means for transmitting the generated voice data, and means for notifying the user when specific emotions or keywords are detected and providing options for intervention.

[1235] System Overview

[1236] Receiving and converting audio data:

[1237] The device detects an incoming call and sends audio data to the server in real time when the call begins. The server uses a speech recognition engine (e.g., Google Speech Recognition) to convert the audio data into text data.

[1238] Text data analysis and response generation:

[1239] The server passes text data to an NLP engine, which analyzes the content and intent of the conversation. Based on the analysis results, it generates a response using a generative AI model (e.g., the OpenAI GPT model). At this time, prompts are used to guide the model in generating the response.

[1240] Emotion recognition by an emotion engine:

[1241] Text and audio data are passed to an emotion engine (e.g., Hugging Face's sentiment-analysis pipeline) to analyze the user's emotions. Based on the recognized emotions, the server adjusts its response.

[1242] Sending responses and notifications:

[1243] The generated audio data is sent to the terminal and then forwarded from the terminal to the customer. Simultaneously, the server detects specific emotions and keywords, and if important call content or emotions are detected, it notifies the user. The user reviews the notification, and if they choose to intervene, the call is seamlessly handed over to them.

[1244] Specific example

[1245] For example, if a customer says, "The video won't play," the system will process it as follows:

[1246] The terminal sends audio data to the server, and the server converts the audio data into text.

[1247] The system analyzes text data and uses a generative AI model to generate an appropriate response ("We apologize, we will investigate").

[1248] The emotion engine detects user dissatisfaction, and the server adjusts its response to be more attentive.

[1249] The server converts the response into audio data and sends it to the customer from the terminal.

[1250] When emotions and keywords are detected, a notification is sent to the user, and they are given the option to intervene.

[1251] Example of a prompt:

[1252] A user said: "The video won't play." How do you respond?

[1253] The system described above enables prompt and appropriate responses to customer inquiries and complaints. Furthermore, by providing responses that take user emotions into account, improved customer satisfaction can be expected.

[1254] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[1255] Step 1:

[1256] The terminal detects an incoming call and receives audio data in real time when the call begins. It then sends the received audio data to the server. Specifically, the terminal uses a microphone to capture audio and transfers the audio data to the server in digital format. In this step, the input is the audio data, and the output is the audio data sent to the server.

[1257] Step 2:

[1258] The server converts the received audio data into text data in real time using a speech recognition engine (e.g., Google Speech Recognition). The server converts the audio to text using digital signal processing and further generates text data for the next processing step. In this step, the input is audio data and the output is text data.

[1259] Step 3:

[1260] The server passes text data to a natural language processing engine for analysis. During the analysis process, it extracts specific intentions and themes and generates data to provide to the AI ​​model that generates appropriate prompt sentences. For example, the keyword "cannot be played" may be detected from the text data. In this step, the input is text data, and the output is the analysis results and prompt sentences for the AI ​​model.

[1261] Step 4:

[1262] The server inputs a prompt message into the generative AI model, which then generates an appropriate response. The generative AI model generates a text-based response based on the provided prompt message. This text-based response contains the solution to the user's question or problem. In this step, the input is the prompt message, and the output is the text data of the generated response.

[1263] Step 5:

[1264] The server passes the generated response text data to a speech synthesis engine (e.g., pyttsx3) to convert it into speech data. The speech synthesis engine converts the text into natural-sounding speech, making it ready to send the response to the customer. In this step, the input is the response text data, and the output is the response speech data.

[1265] Step 6:

[1266] The server passes text and audio data to the emotion engine (e.g., Hugging Face's sentiment-analysis pipeline) to analyze the user's emotions. This analysis identifies the user's positive, negative, or neutral emotions and may reflect them in the response. In this step, the input is text and audio data, and the output is the emotion analysis result.

[1267] Step 7:

[1268] The server further refines the generated response based on the sentiment analysis results, providing it in a way that is appropriate to the user's emotions. For example, if the user is angry, the response is adjusted to be more polite. In this step, the input is the sentiment analysis results and the generated response text data, and the output is the adjusted response text data.

[1269] Step 8:

[1270] The server sends the finalized response audio data to the terminal, which then sends it back to the caller. The terminal plays the audio back to the customer in real time, conveying the content of the response. In this step, the input is the adjusted response audio data, and the output is the response to the customer.

[1271] Step 9:

[1272] If specific emotions and keywords are detected, the server sends a notification to the user informing them that intervention is needed. The user reviews the notification and chooses whether to intervene in the call if necessary. In this step, the inputs are the emotion analysis results and keyword detection results, and the outputs are the notification to the user and the choice of intervention.

[1273] The specific processing unit 290 transmits the result of the specific processing to the headset terminal 314. In the headset terminal 314, the control unit 46A causes the speaker 240 and display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[1274] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1275] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and specific processing may also be performed by the headset terminal 314.

[1276] [Fourth Embodiment]

[1277] Figure 7 shows an example of the configuration of the data processing system 410 according to the fourth embodiment.

[1278] As shown in Figure 7, the 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.

[1279] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1280] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a controlled object 443. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and controlled object 443 are also connected to the bus 52.

[1281] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[1282] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[1283] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[1284] The controlled object 443 includes a display device, LEDs in the eyes, and motors that drive 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 robot 414's emotions can be expressed by controlling these motors. Furthermore, the robot 414's facial expressions can also be expressed by controlling the illumination state of the LEDs in its eyes.

[1285] Figure 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Figure 8, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[1286] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[1287] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1288] In robot 414, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[1289] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1290] The system according to the present invention receives voice data, converts it to text data in real time, analyzes it, generates an appropriate response, and transmits it as voice data. It also detects specific keywords and notifies the user, enabling seamless switching so that the user can intervene as needed.

[1291] System-wide configuration

[1292] The system consists of the following elements:

[1293] 1. Terminal

[1294] 2. Server

[1295] 3. User

[1296] Receiving and converting audio data

[1297] The device detects an incoming call and, as soon as the call begins, sends the audio data to the server in real time.

[1298] The server passes the received audio data to the speech recognition engine, which converts it into text data in real time.

[1299] Text data analysis and response generation

[1300] The server uses a natural language processing (NLP) engine to analyze text data and understand its content.

[1301] Based on the analysis results, the server automatically generates an appropriate response using a generation AI. This response is then generated again in text format.

[1302] The server passes the text-based response to the speech synthesis engine, which converts it into speech data.

[1303] Sending a response and notification

[1304] The server sends the generated voice data to the terminal, which then relays it to the person on the other end of the line.

[1305] The server continuously monitors the call content and notifies the user via the terminal if it detects specific keywords.

[1306] Seamless user intervention

[1307] Users can check notifications and, if important call content is detected, intervene in the call themselves.

[1308] The device accepts user intervention, allowing the user to speak directly with the person they are calling.

[1309] Specific example

[1310] 1. Handling sales calls

[1311] A salesman makes a phone call and says, "Excuse me for bothering you while you're busy, this is a salesperson from AA Insurance."

[1312] The device transmits voice data to the server in real time.

[1313] The server converts the voice data into text and recognizes it as "This is a salesperson from AA Insurance."

[1314] The server uses a generation AI to generate the response, "I'm not interested, thank you," converts it into audio data, and sends it to the terminal.

[1315] The device responds to the salesman by playing this audio data.

[1316] 2. Important notifications and user intervention

[1317] The salesman continued, "Would you mind if I sent you some materials?"

[1318] The server detects specific keywords such as "send documents" and notifies the user via the terminal.

[1319] The user checks the notification and taps the screen to intervene in the call, directly responding, "I don't need to send any documents."

[1320] Thus, the present invention is a system that automates telephone answering while allowing user intervention in important situations, thereby achieving efficient and flexible responses.

[1321] Next, we can describe the details of the processing steps.

[1322] The following describes the processing flow.

[1323] Step 1:

[1324] The device detects an incoming call and collects the call audio as soon as the call begins. The collected audio data is transmitted to the server in real time.

[1325] Step 2:

[1326] The server passes the received audio data to the speech recognition engine, which converts the audio data into text data. This conversion is performed in real time and processed sequentially.

[1327] Step 3:

[1328] The server passes the text data to a natural language processing (NLP) engine for content analysis. The analysis identifies the purpose and topic of the call.

[1329] Step 4:

[1330] The server then uses the generation AI to create a response based on the analysis results. This response is generated in text format and designed to be an appropriate reply to the conversation.

[1331] Step 5:

[1332] The server passes the generated text response to the speech synthesis engine, which converts it into speech data. The converted speech data is then generated again in real time.

[1333] Step 6:

[1334] The server sends the audio data to the terminal. The terminal decodes the received audio data and plays it back to the caller at the appropriate time.

[1335] Step 7:

[1336] The server continues to monitor the call content to check for the inclusion of specific keywords or phrases. For example, it may detect specific keywords such as "important matter" or "sending documents."

[1337] Step 8:

[1338] When a specific keyword is detected, the server sends a notification to the terminal, informing the user that intervention is necessary.

[1339] Step 9:

[1340] The device displays a notification to the user, alerting them with audio and visuals. It then prompts the user to choose whether or not to take action.

[1341] Step 10:

[1342] The user checks the notification and, if important call content is detected, intervenes in the call at their own discretion. The device accepts the user's intervention and allows the user to speak directly with the other party.

[1343] Step 11:

[1344] Once the call ends, the device sends the audio recording and text data to the server. The server stores this data in a database for later reference.

[1345] In this way, the system seamlessly integrates real-time voice responses with user intervention, enabling effective telephone handling.

[1346] (Example 1)

[1347] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1348] Conventional voice response systems were unable to adequately generate responses or detect keywords in real time, making it difficult to respond immediately to important call content. Furthermore, the inability to seamlessly integrate automated responses with user intervention made flexible and efficient call handling challenging.

[1349] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[1350] In this invention, the server includes means for receiving voice data and converting it into text data in real time; means for analyzing the text data and generating an appropriate response using a generative AI model; means for converting the generated text data of the response into voice data; means for transmitting the generated voice data of the response; means for detecting specific keywords and notifying the user; and means for allowing the user to acknowledge the notification and intervene in the call. This enables real-time text conversion and response generation, keyword detection and user notification, and seamless user intervention.

[1351] "Audio data" refers to audio signals obtained through phone calls, recordings, etc., represented in digital format.

[1352] "Real-time" refers to data processing and responses occurring immediately, with results reflected without delay.

[1353] "Text data" refers to audio data converted into written information, expressed in a format that is understandable to humans.

[1354] A "generative AI model" refers to a technology that uses artificial intelligence algorithms to generate appropriate responses or data based on given input.

[1355] A "prompt" is a series of text inputs to a generative AI model, which the AI ​​uses to generate a response.

[1356] "Analysis" is a means of processing text data to understand its content and structure, and a process for deriving appropriate responses or actions.

[1357] A "keyword" is an important word or phrase contained within text data, and its detection triggers a specific action or response.

[1358] A "user" is a person who uses the system and has the authority to intervene in the content of a call.

[1359] A "notification" is an informational alert sent from the system to the user, intended to draw the user's attention when a specific keyword is detected.

[1360] "Intervening in a call" refers to the action of a user switching from an automated system response to a manual response and starting a direct conversation with the other party.

[1361] A "speech recognition engine" is software or hardware that analyzes speech data and converts it into text data.

[1362] A "speech synthesis engine" is software or hardware that analyzes text data and converts it into speech data.

[1363] "Communication methods" refer to technologies used to send and receive voice and text data between a server and a terminal, and include the Internet Protocol and other communication protocols.

[1364] A specific embodiment of the system according to the present invention is shown below. This system receives voice data, converts it into text data in real time, performs subsequent analysis and response generation, and enables notifications and call interventions to the user.

[1365] Receiving and converting audio data

[1366] The terminal detects an incoming call and captures audio data as soon as the call begins. The call audio is transmitted to the server in real time. WebSocket and HTTP streaming technologies are used as secure communication methods for transmitting the audio data.

[1367] The server converts the received audio data into text data using the Google Cloud Speech-to-Text API or Amazon Transcribe. The converted text data is extracted as a string that accurately represents the spoken content.

[1368] Text data analysis and response generation

[1369] The server uses Google Cloud Natural Language API and Microsoft Azure Text Analytics to analyze text data and understand its content. The information extracted through analysis includes grammatical structure, meaning, and important keywords.

[1370] Based on the analysis results, the server automatically generates an appropriate response using a generative AI model (e.g., GPT-3). The generative AI model used here has advanced natural language generation capabilities and generates the optimal response from the given prompt sentence.

[1371] The generated response is then converted back into text format and then into speech data using the Google Text-to-Speech API or Amazon Polly.

[1372] Sending a response and notification

[1373] The server sends the generated audio data to the terminal, which then relays that audio data to the person on the other end of the call. This enables real-time responses.

[1374] A server that continuously monitors call content will notify the user via the terminal if it detects a specific keyword, such as the phrase "send documents." The notification system can display the notification on the terminal screen or alert the user with vibration or sound.

[1375] Seamless user intervention

[1376] Upon receiving a notification, the user can review the notification displayed on their device screen and decide whether or not to intervene in the call. If the user chooses to intervene, tapping the notification on their device will switch the system from automatic response to manual response, allowing the user to speak directly with the caller.

[1377] Specific example

[1378] 1. Handling sales calls

[1379] A salesman makes a phone call and says, "Excuse me for bothering you while you're busy, this is a salesperson from AA Insurance."

[1380] The device transmits voice data to the server in real time.

[1381] The server converts the voice data into text and recognizes it as "This is a salesperson from AA Insurance."

[1382] The server uses a generation AI to generate the response, "I'm not interested, thank you," converts it into audio data, and sends it to the terminal.

[1383] The device responds to the salesman by playing this audio data.

[1384] 2. Important notifications and user intervention

[1385] The salesman continued, "Would you mind if I sent you some materials?"

[1386] The server detects specific keywords such as "send documents" and notifies the user via the terminal.

[1387] The user checks the notification, taps the screen to intervene in the call, and directly responds, "I don't need to send any documents."

[1388] Example of a prompt

[1389] "If a salesperson is soliciting insurance and includes the phrase 'send me some materials,' the user should be notified. The response should generate 'I do not need any materials.'"

[1390] This system enables efficient and flexible call handling by transcribing audio in real time, generating appropriate responses, and allowing user intervention at key points.

[1391] The flow of the specific processing in Example 1 will be explained using Figure 11.

[1392] Step 1:

[1393] The device detects an incoming call and captures audio data as soon as the call begins. The audio data is sent to the server in real time. Specifically, it receives the phone's audio in real time as input and sends that audio data to the server via the network as output. In practice, the smartphone outputs a log message saying "Audio capture started" and immediately begins sending the audio data.

[1394] Step 2:

[1395] The server passes the received audio data to a speech recognition engine (e.g., Google Cloud Speech-to-Text API) and converts it into text data in real time. It takes audio data as input and generates text data as output. Specifically, the server outputs a log message saying "Audio data converted to text."

[1396] Step 3:

[1397] The server uses a natural language processing (NLP) engine (e.g., Google Cloud Natural Language API) to analyze text data. It receives text data as input and generates analysis results as output. Analysis includes entity extraction and sentiment analysis. Specifically, the server logs "Text analysis complete: Intent is sales."

[1398] Step 4:

[1399] The server generates an appropriate response based on the analysis results using a generative AI model (e.g., GPT-3). It receives the analysis results and a prompt as input and generates the appropriate response text as output. Specifically, the server logs "Response generation complete." An example of a prompt is: "If a salesman is soliciting insurance and includes the phrase 'send materials,' the user should be notified. The response should generate 'No materials are needed.'"

[1400] Step 5:

[1401] The server passes the generated response text to a text-to-speech engine (e.g., Google Text-to-Speech API) to convert it into speech data. It takes response text as input and generates speech data as output. Specifically, the server logs "Text-to-speech complete".

[1402] Step 6:

[1403] The server sends the generated audio data to the terminal, and the terminal transmits that audio data to the person it is talking to. It receives audio data as input and plays it back to the person it is talking to as output. Specifically, the terminal log displays "Playing response audio".

[1404] Step 7:

[1405] The server continuously monitors the call content and, if it detects a specific keyword (e.g., "send documents"), it notifies the user via the terminal. It receives real-time text data as input and generates a notification as output. Specifically, the server log records "Keyword detected: Send documents" and sends a notification to the terminal.

[1406] Step 8:

[1407] The user can check notifications and, if important call content is detected, intervene in the call themselves. The device receives notifications as input and intervenes in the call as output. Specifically, the user taps the notification on their device and directly responds with "No documents are needed." The device then logs "User intervened in the call."

[1408] (Application Example 1)

[1409] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1410] Traditional content distribution services have faced challenges in providing personalized recommendations based on user preferences and interests, making it difficult to respond quickly and accurately to the information and entertainment users seek. Furthermore, the lack of systems that use AI-generated models to respond to requests has made it difficult to obtain the information users need in real time.

[1411] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[1412] In this invention, the server includes means for receiving voice data and converting it into text data in real time; means for analyzing the text data, generating a response using a generative AI model based on the analysis results, and converting the response into voice data; means for transmitting the generated voice data of the response; means for obtaining a request from the user in voice data and sending a prompt sentence to the generative AI model based on the request; and means for suggesting appropriate content to the user based on the generated prompt. This makes it possible to recommend the most suitable content to the user in real time.

[1413] "Audio data" refers to data obtained by converting audio into a digital format.

[1414] "Real-time" refers to processing data instantly and without delay.

[1415] "Text data" refers to digital data obtained by converting audio data into written text.

[1416] A "generative AI model" refers to an artificial intelligence model that generates appropriate responses based on user input data.

[1417] A "prompt" is a portion of the text data input to a generating AI model, and it functions as an instruction or question for the model.

[1418] "Analysis" refers to the process of understanding text data and extracting its meaning.

[1419] "Content" is a general term for information and entertainment that users consume, such as movies, music, and news.

[1420] "Suggestions" refer to the act of recommending appropriate content based on the user's preferences and requests.

[1421] This invention relates to a system that converts audio data into text data in real time, analyzes the text data to generate a response using a generative AI model, and transmits that response back as audio data. Furthermore, it also includes a function to generate prompt sentences for the generative AI model based on user requests and suggest appropriate content.

[1422] System Configuration

[1423] The system consists of the following elements:

[1424] 1. Terminal

[1425] An input device for receiving audio data (e.g., a microphone)

[1426] An output device for playing audio data (e.g., a speaker)

[1427] 2. Server

[1428] Speech recognition engine

[1429] Natural Language Processing (NLP) Engine

[1430] Generative AI models (e.g., GPT-3)

[1431] Speech synthesis engine

[1432] 3. User

[1433] Person making a request by voice

[1434] Receiving and converting audio data

[1435] The terminal receives the user's voice data and sends it to the server in real time. The server uses a speech recognition engine to convert the voice data into text data.

[1436] Text data analysis and response generation

[1437] The server analyzes the converted text data using a natural language processing (NLP) engine to understand its content. Based on the analysis results, it uses a generative AI model to generate an appropriate response, which is then generated again in text format. The generated text is then converted into speech data by a speech synthesis engine.

[1438] Sending a response and suggesting content

[1439] The generated audio data is sent to the device and transmitted to the user through the device's speaker. Furthermore, a request is obtained from the user via the audio data, and a prompt is sent to the generating AI model based on that request. Based on the generated prompt, appropriate content is suggested to the user.

[1440] Specific example

[1441] For example, if a user requests "Tell me your recommendations for the latest movies," the following prompt message will be generated.

[1442] Examples of prompts to input into a generative AI model:

[1443] User request: Please recommend some of the latest movies.

[1444] Appropriate response:

[1445] Based on this prompt, the server sends information to the AI ​​model that generates the appropriate response text. For example, a response such as "As a recent movie recommendation, we suggest the action film XXX" is generated, converted into audio data, and delivered to the user.

[1446] This allows users to receive real-time responses to their requests and to have content that interests them suggested quickly and accurately.

[1447] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[1448] Step 1:

[1449] The device receives the user's voice data.

[1450] Input: User's voice data

[1451] Operation: The microphone captures the user's voice and converts the audio data into a digital format.

[1452] Output: Digital audio data

[1453] Step 2:

[1454] The terminal sends the received digital audio data to the server.

[1455] Input: Digital audio data

[1456] Operation: Transmits audio data to the server in real time via the network interface.

[1457] Output: Audio data sent to the server

[1458] Step 3:

[1459] The server uses a speech recognition engine to convert the audio data into text data.

[1460] Input: Audio data

[1461] Operation: Analyzes audio data using a speech recognition engine (e.g., Google Speech Recognition) and converts it to text.

[1462] Output: Text data

[1463] Step 4:

[1464] The server analyzes the text data using a natural language processing (NLP) engine and understands its meaning.

[1465] Input: Text data

[1466] Operation: Uses a natural language processing (NLP) engine to extract meaning from text data.

[1467] Output: Analysis results (user request details)

[1468] Step 5:

[1469] Based on the analysis results, the server generates a response using an AI model.

[1470] Input: Analysis results

[1471] Operation: A prompt is sent to a generative AI model (e.g., GPT-3), and a response text is generated based on it. An example prompt is "User request: Tell me your recommendations for the latest movies. Appropriate response:"

[1472] Output: Response text data

[1473] Step 6:

[1474] The server converts the generated response text into speech data using a speech synthesis engine.

[1475] Input: Response text data

[1476] Operation: Uses a speech synthesis engine (e.g., pyttsx3) to convert response text into speech data.

[1477] Output: Response voice data

[1478] Step 7:

[1479] The server sends the generated response voice data to the terminal.

[1480] Input: Response voice data

[1481] Operation: Transmits voice data to the terminal in real time via the network interface.

[1482] Output: Response voice data sent to the terminal

[1483] Step 8:

[1484] The device plays back the received voice response data and conveys it to the user.

[1485] Input: Response voice data

[1486] Operation: Plays audio data through the speaker.

[1487] Output: The user receives a voice response.

[1488] Step 9:

[1489] If a new request comes from the user, the same process is repeated.

[1490] Input: New voice request

[1491] Action: Repeat the process from step 1.

[1492] Output: New response to the user

[1493] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[1494] The system related to this invention automates telephone answering and allows for seamless user intervention as needed. Furthermore, this system incorporates an emotion engine that recognizes the user's emotions and adjusts its response accordingly.

[1495] System-wide configuration

[1496] The system consists of the following elements:

[1497] 1. Terminal

[1498] 2. Server

[1499] 3. User

[1500] 4. Emotional Engine

[1501] Receiving and converting audio data

[1502] The device detects an incoming call and sends audio data to the server in real time when the call begins. The audio data is the content of the call and is sent to the server sequentially.

[1503] Text data analysis and response generation

[1504] The server receives the audio data and converts it into text data in real time using a speech recognition engine.

[1505] The server passes the text data to the NLP engine, which analyzes the content and intent of the conversation. This clarifies the specific topic and purpose of the call.

[1506] Based on the analysis results, the server uses a generative AI to generate an appropriate response in text format. This process builds an automated response system for calls.

[1507] The server passes the generated text response to the speech synthesis engine, which converts it into speech data in real time.

[1508] Emotion recognition by an emotion engine

[1509] The server passes text and audio data to the emotion engine, which analyzes the user's emotions. For example, emotions can be inferred from the tone of voice and word choice.

[1510] Based on the emotions recognized by the emotion engine, the server adjusts its response. If the user is angry, a more careful response will be generated.

[1511] The server detects specific emotions and keywords and sends a notification to the user based on them. In this case, it is determined that the user needs to intervene.

[1512] Sending a response and notification

[1513] The server sends the generated audio data to the terminal, which then sends it back to the person on the other end of the call. Simultaneously, if important conversation content or emotions are detected, the user is notified.

[1514] The device displays a notification to the user and offers them an option to intervene. If the user chooses to intervene, the call is seamlessly handed over to the user.

[1515] Specific example

[1516] 1. Examples of handling customer complaints

[1517] A customer calls to complain: "The item I ordered hasn't arrived yet!"

[1518] The device sends the audio data to the server.

[1519] The server converts the audio data into text and analyzes it using an NLP engine.

[1520] The server detects keywords such as "order," "not received," and "dissatisfaction," and uses a generation AI to generate a response such as "We apologize, we will check on this immediately."

[1521] The emotion engine detects an angry tone and carefully adjusts its response accordingly.

[1522] The server sends the response to the customer via the terminal.

[1523] The server sends a notification to the user based on sentiment data and keywords. It displays a message saying, "We are handling a complaint and intervention is required."

[1524] The user reviews the notification and selects intervention. The system responds directly to the customer with, "We will investigate immediately, please wait."

[1525] In this way, the system of the present invention incorporates emotion recognition functionality to provide appropriate responses based on the user's emotions, and allows the user to intervene as needed. This enables improved efficiency and quality in telephone communication.

[1526] The following describes the processing flow.

[1527] Step 1:

[1528] The device detects incoming calls and collects call audio as soon as the call begins. The collected audio data is sent to the server in real time.

[1529] Step 2:

[1530] The server passes the received audio data to the speech recognition engine, which converts the audio data into text data. This conversion is performed in real time and processed sequentially.

[1531] Step 3:

[1532] The server then passes the text data to a natural language processing (NLP) engine for analysis. The analysis results identify the purpose and topic of the call.

[1533] Step 4:

[1534] The server passes the analyzed text data to the emotion engine, which analyzes the user's emotions. The emotion engine identifies the user's emotions based on their tone of voice and word choice.

[1535] Step 5:

[1536] The server integrates the results from the emotion engine and the NLP engine's analysis, and has the generative AI create an appropriate response. This response is generated in text format and is designed to be an appropriate reply in a conversation.

[1537] Step 6:

[1538] The server passes the generated text response to the speech synthesis engine, which converts it into speech data. The generated speech data is then generated again in real time.

[1539] Step 7:

[1540] The server sends the generated voice data to the terminal, which then sends it to the caller. This is where the response is conveyed to the other party.

[1541] Step 8:

[1542] The server continues to monitor the call content and sends notifications to the user if specific keywords or emotions are detected.

[1543] Step 9:

[1544] The device displays notifications to the user, alerting them with audio and visuals. These notifications may include messages such as, "A customer is expressing dissatisfaction. Intervention is needed."

[1545] Step 10:

[1546] The user reviews the notification and, if important call content or emotions are detected, intervenes in the call at their own discretion. The device accepts the user's intervention.

[1547] Step 11:

[1548] The user chooses to intervene and begins a direct conversation with the person on the other end of the line. This allows for a seamless integration of human judgment into responses, even when emotional responses are required.

[1549] Step 12:

[1550] Once the call ends, the device sends the audio recording and text data to the server. The server stores this data in a database for later reference.

[1551] By combining this with an emotion engine, it becomes possible to automate telephone answering while still providing flexible responses based on the user's emotions.

[1552] (Example 2)

[1553] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1554] Traditional telephone answering systems often relied on manual processes for analyzing call content and generating appropriate responses, resulting in challenges in response quality and efficiency. Furthermore, they struggled to respond in a way that considered user emotions, hindering responses that required quick and appropriate action, particularly in situations like complaint handling. Additionally, delays in detecting specific key keywords or emotions could lead to delayed user intervention and inadequate responses.

[1555] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[1556] In this invention, the server includes means for detecting an incoming call and transmitting voice data in real time when the call begins; means for receiving the voice data and converting it into text data in real time using recognition means; means for passing the text data to analysis means and creating a response based on the analysis results using generation means; means for passing the text data of the generated response to synthesis means and converting it into voice data; and means for transmitting the generated voice data. This enables real-time analysis of call content and automatic generation of appropriate responses.

[1557] "Means for detecting incoming phone calls" refers to devices or software that identify call requests received via telephone lines or internet communication.

[1558] "Means of transmitting voice data in real time" refers to devices or software that have the function of instantly processing voice information spoken during a call and transmitting it to other devices or systems via a network.

[1559] "Means for receiving audio data and converting it into text data in real time using recognition means" refers to devices or software that have the function of instantly converting received audio information into text data using a speech recognition engine.

[1560] "Means for passing text data to an analysis means and creating a response based on the analysis results using a generation means" refers to devices or software that send text data to a natural language processing engine and create an appropriate response sentence using a generation AI or similar based on the analysis results.

[1561] "Means for passing the generated response text data to a synthesis means and converting it into audio data" refers to devices or software that have the function of sending a text-based response to a speech synthesis engine and converting it into a format that can be played back as audio information.

[1562] "Means for transmitting generated audio data" refers to devices or software that have the function of transmitting audio data generated by a speech synthesis engine to a designated terminal or receiving device.

[1563] "Means for passing text data and audio data to emotion recognition means and analyzing the user's emotions" refers to devices or software that have the function of determining emotions from text and audio using an analysis engine and identifying the speaker's emotional state.

[1564] "Means of adjusting response content based on emotions" refers to devices or software that have the function of modifying responses to appropriate tone and wording based on emotional information identified by an analysis engine.

[1565] "Means of notifying a user when specific keywords are detected, allowing the user to intervene in a call" refers to devices or software that have the function of notifying an administrator or operator when pre-set important keywords or emotional states are detected, providing them with an opportunity to intervene in the situation.

[1566] The system related to this invention automates telephone answering, allowing user intervention as needed. Furthermore, the system incorporates an emotion engine that recognizes the user's emotions and adjusts its response accordingly. This system includes the following elements:

[1567] 1. Terminal

[1568] 2. Server

[1569] 3. User

[1570] 4. Emotional Engine

[1571] Receiving and converting audio data

[1572] The device detects an incoming call and, once the call begins, sends audio data to the server in real time. This audio data, containing the conversation content, is transmitted to the server continuously. Specifically, smartphones and landline phones fulfill this role.

[1573] Text data analysis and response generation

[1574] The server receives the audio data and converts it into text data in real time using a speech recognition engine (e.g., Google Cloud Speech-to-Text API). The converted text data is then passed to a natural language processing engine (e.g., Google Cloud Natural Language API) to analyze the content and intent of the conversation.

[1575] Next, based on the analysis results, a generative AI (e.g., OpenAI's GPT-4) is used to generate an appropriate response in text format. The generated text response is then passed to a speech synthesis engine (e.g., Microsoft Azure Text-to-Speech) and converted into speech data in real time.

[1576] Emotion recognition and response adjustment by an emotion engine

[1577] The server passes text and audio data to an emotion engine (e.g., IBM Watson Tone Analyzer) to analyze the user's emotions. Based on the analysis, the response is adjusted. For example, if the user is angry, the response will be more careful and polite.

[1578] Sending a response and notification

[1579] The server sends the generated audio data to the terminal, which then sends it back to the caller. Simultaneously, if important call content or emotions are detected, the user is notified. The notification might say something like, "We are handling a complaint. Intervention is required."

[1580] User intervention

[1581] The device displays a notification to the user and offers an option to intervene. If the user chooses to intervene, the call is seamlessly handed over to the user. The user can then acknowledge the notification and respond directly.

[1582] Specific example

[1583] For example, consider a scenario where a customer calls to complain, saying, "The item I ordered hasn't arrived yet!"

[1584] The device sends the audio data to the server.

[1585] The server converts the audio data into text and analyzes it using an NLP engine.

[1586] The server detects keywords such as "order," "not received," and "dissatisfaction," and uses a generation AI to generate a response such as "We apologize, we will check on this immediately."

[1587] The emotion engine detects an angry tone and carefully adjusts its response accordingly.

[1588] The server sends the response to the customer via the terminal.

[1589] Based on sentiment data and keywords, the server sends a notification to the user stating, "We are handling a complaint; intervention is required."

[1590] The user reviews the notification and chooses to intervene. The system responds directly to the customer with, "We will investigate immediately, please wait."

[1591] Example of a prompt

[1592] "We have received an inquiry from a customer regarding a non-delivery of their product, and their message is filled with anger. Please prepare an appropriate response."

[1593] "The customer is dissatisfied with the delay in their order. Please generate a polite response to improve this situation."

[1594] In this way, the system of the present invention incorporates emotion recognition functionality to provide appropriate responses based on the user's emotions, and allows the user to intervene as needed. This enables improved efficiency and quality in telephone communication.

[1595] The flow of the specific processing in Example 2 will be explained using Figure 13.

[1596] Step 1:

[1597] The device detects incoming calls and, once a call begins, sends audio data to the server in real time. Specifically, the device's phone function detects the start of a call, converts the audio picked up by the microphone into digital audio data, and sends it to the server via the internet.

[1598] Input: Incoming phone call

[1599] Output: Audio data transmitted in real time

[1600] Step 2:

[1601] The server receives audio data and converts it into text data in real time using a speech recognition engine (e.g., Google Cloud Speech-to-Text API). Specifically, the speech recognition engine analyzes the audio signal and converts the words used in the conversation into text format.

[1602] Input: Audio data transmitted in real time

[1603] Output: Text data

[1604] Step 3:

[1605] The server passes the converted text data to a natural language processing engine (e.g., Google Cloud Natural Language API) to analyze the conversation. Specifically, the natural language processing engine extracts meaning from the text data and identifies important keywords and topics.

[1606] Input: Text data converted by the speech recognition engine

[1607] Output: Analyzed keyword and topic information

[1608] Step 4:

[1609] The server uses a generative AI model (e.g., OpenAI's GPT-4) based on the analysis results to generate an appropriate response in text format. It utilizes prompts to adjust the generated response as needed.

[1610] Input: Analyzed keyword and topic information

[1611] Output: Generated text response

[1612] Example prompt:

[1613] "We have received an inquiry from a customer regarding a non-delivery of their product, and their message is filled with anger. Please prepare an appropriate response."

[1614] Step 5:

[1615] The server passes the generated text response to a text-to-speech engine (e.g., Microsoft Azure Text-to-Speech), which converts it into speech data in real time. Specifically, the text-to-speech engine converts the text into speech, creating natural and fluent speech output.

[1616] Input: Generated text response

[1617] Output: Generated audio data

[1618] Step 6:

[1619] The server passes the voice and text data to an emotion recognition engine (e.g., IBM Watson Tone Analyzer) to analyze the user's emotions. Specifically, the emotion recognition engine identifies emotions from the tone and word choice of the voice and then identifies areas that need adjustment based on that.

[1620] Input: Generated text response and audio data

[1621] Output: Emotion analysis results

[1622] Step 7:

[1623] The server adjusts the generated response based on the recognized emotion. For example, if anger is detected, the response will be made more careful and polite.

[1624] Input: Sentiment analysis results

[1625] Output: Adjusted text response and audio data

[1626] Step 8:

[1627] The server sends the generated audio data to the terminal, and the terminal sends it back to the person on the other end of the line. Specifically, the server sends audio data to the terminal in real time, and the terminal transmits that data to the person on the other end of the line through its audio output device (speaker).

[1628] Input: Adjusted audio data

[1629] Output: Transmitted audio data

[1630] Step 9:

[1631] Based on text data and sentiment analysis results, the server sends notifications to the user when specific keywords or important emotions are detected. These notifications may include alerts such as, "We are handling a complaint. Intervention is required."

[1632] Input: Specific keywords and sentiment analysis results

[1633] Output: User statement notification

[1634] Step 10:

[1635] The device displays a notification to the user and offers an option to intervene. If the user chooses to intervene, the call is seamlessly handed over to the user. Specifically, the user acknowledges the notification and responds directly to the call if necessary.

[1636] Input: User Statement Notification

[1637] Output: User intervention to handle calls

[1638] (Application Example 2)

[1639] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1640] Traditional telephone answering systems failed to consider customer emotions, and inappropriate responses easily led to customer dissatisfaction. Furthermore, automated answering systems could only handle a limited range of questions, requiring human intervention for complex inquiries and complaints, and this intervention was sometimes not timely. In addition, content delivery services require prompt and appropriate responses to inquiries to improve the user experience.

[1641] The specific processing performed by the specific 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 receiving voice data and converting it into text data in real time, means for analyzing the text data and generating a response using a generation AI model based on the analysis results, means for converting the generated response into voice data, means for analyzing the user's emotions from the text data and voice data using an emotion engine, means for transmitting the voice data of the generated response, and means for notifying the user when specific emotions and keywords are detected and providing an option for intervention. This provides an appropriate automated response that takes the user's emotions into consideration, and allows for human intervention at the appropriate time as needed.

[1642] "Audio data" refers to information that represents telephone calls or recorded audio in digital format.

[1643] "Text data" refers to character information generated by analyzing audio data and using speech recognition technology.

[1644] A "generative AI model" is a type of artificial intelligence that automatically generates appropriate responses to specific inputs (prompt sentences) based on natural language processing.

[1645] A "prompt statement" is an input statement used to generate a response for a generative AI model.

[1646] An "emotion engine" is an algorithm or system that analyzes voice and text data to recognize the user's emotions (such as joy, anger, or sadness) contained within it.

[1647] "Analysis results" refer to conclusions or information obtained through analysis of text data using natural language processing or sentiment engines.

[1648] "Response" refers to the reply or action generated by a generative AI model based on its analysis results.

[1649] A "notification" is an alert or message sent from a system to a user under specific conditions (for example, detection of the user's emotions or keywords).

[1650] "Intervention" refers to the act of a user directly participating in a call or response when the system cannot handle the situation through automated responses or under specific conditions.

[1651] "Means" refer to equipment, algorithms, systems, methods, etc., used to achieve a specific purpose.

[1652] The present invention relates to a system that automates telephone answering and recognizes user emotions to adjust responses. The system comprises means for receiving voice data and converting it into text data in real time, means for analyzing the text data and generating responses using a generative AI model, means for converting the generated responses into voice data, means for analyzing the user's emotions using an emotion engine, means for transmitting the generated voice data, and means for notifying the user when specific emotions or keywords are detected and providing options for intervention.

[1653] System Overview

[1654] Receiving and converting audio data:

[1655] The device detects an incoming call and sends audio data to the server in real time when the call begins. The server uses a speech recognition engine (e.g., Google Speech Recognition) to convert the audio data into text data.

[1656] Text data analysis and response generation:

[1657] The server passes text data to an NLP engine, which analyzes the content and intent of the conversation. Based on the analysis results, it generates a response using a generative AI model (e.g., the OpenAI GPT model). At this time, prompts are used to guide the model in generating the response.

[1658] Emotion recognition by an emotion engine:

[1659] Text and audio data are passed to an emotion engine (e.g., Hugging Face's sentiment-analysis pipeline) to analyze the user's emotions. Based on the recognized emotions, the server adjusts its response.

[1660] Sending responses and notifications:

[1661] The generated audio data is sent to the terminal and then forwarded from the terminal to the customer. Simultaneously, the server detects specific emotions and keywords, and if important call content or emotions are detected, it notifies the user. The user reviews the notification, and if they choose to intervene, the call is seamlessly handed over to them.

[1662] Specific example

[1663] For example, if a customer says, "The video won't play," the system will process it as follows:

[1664] The terminal sends audio data to the server, and the server converts the audio data into text.

[1665] The system analyzes text data and uses a generative AI model to generate an appropriate response ("We apologize, we will investigate").

[1666] The emotion engine detects user dissatisfaction, and the server adjusts its response to be more attentive.

[1667] The server converts the response into audio data and sends it to the customer from the terminal.

[1668] When emotions and keywords are detected, a notification is sent to the user, and they are given the option to intervene.

[1669] Example of a prompt:

[1670] A user said: "The video won't play." How do you respond?

[1671] The system described above enables prompt and appropriate responses to customer inquiries and complaints. Furthermore, by providing responses that take user emotions into account, improved customer satisfaction can be expected.

[1672] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[1673] Step 1:

[1674] The terminal detects an incoming call and receives audio data in real time when the call begins. It then sends the received audio data to the server. Specifically, the terminal uses a microphone to capture audio and transfers the audio data to the server in digital format. In this step, the input is the audio data, and the output is the audio data sent to the server.

[1675] Step 2:

[1676] The server converts the received audio data into text data in real time using a speech recognition engine (e.g., Google Speech Recognition). The server converts the audio to text using digital signal processing and further generates text data for the next processing step. In this step, the input is audio data and the output is text data.

[1677] Step 3:

[1678] The server passes text data to a natural language processing engine for analysis. During the analysis process, it extracts specific intentions and themes and generates data to provide to the AI ​​model that generates appropriate prompt sentences. For example, the keyword "cannot be played" may be detected from the text data. In this step, the input is text data, and the output is the analysis results and prompt sentences for the AI ​​model.

[1679] Step 4:

[1680] The server inputs a prompt message into the generative AI model, which then generates an appropriate response. The generative AI model generates a text-based response based on the provided prompt message. This text-based response contains the solution to the user's question or problem. In this step, the input is the prompt message, and the output is the text data of the generated response.

[1681] Step 5:

[1682] The server passes the generated response text data to a speech synthesis engine (e.g., pyttsx3) to convert it into speech data. The speech synthesis engine converts the text into natural-sounding speech, making it ready to send the response to the customer. In this step, the input is the response text data, and the output is the response speech data.

[1683] Step 6:

[1684] The server passes text and audio data to the emotion engine (e.g., Hugging Face's sentiment-analysis pipeline) to analyze the user's emotions. This analysis identifies the user's positive, negative, or neutral emotions and may reflect them in the response. In this step, the input is text and audio data, and the output is the emotion analysis result.

[1685] Step 7:

[1686] The server further refines the generated response based on the sentiment analysis results, providing it in a way that is appropriate to the user's emotions. For example, if the user is angry, the response is adjusted to be more polite. In this step, the input is the sentiment analysis results and the generated response text data, and the output is the adjusted response text data.

[1687] Step 8:

[1688] The server sends the finalized response audio data to the terminal, which then sends it back to the caller. The terminal plays the audio back to the customer in real time, conveying the content of the response. In this step, the input is the adjusted response audio data, and the output is the response to the customer.

[1689] Step 9:

[1690] If specific emotions and keywords are detected, the server sends a notification to the user informing them that intervention is needed. The user reviews the notification and chooses whether to intervene in the call if necessary. In this step, the inputs are the emotion analysis results and keyword detection results, and the outputs are the notification to the user and the choice of intervention.

[1691] 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 controlled object 443 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[1692] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1693] 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 this disclosure is not limited thereto, and the specific processing may also be performed by the robot 414.

[1694] Furthermore, the emotion identification model 59, acting 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 a specific mapping, which is an emotion map (see Figure 9). Similarly, the emotion identification model 59 may also determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1695] Figure 9 shows an emotion map 400 in which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. The closer to the center of the concentric circles, the more primitive the emotions are located. Further out of the concentric circles, emotions representing states and actions arising from mental states are located. Emotion is a concept that includes feelings and mental states. On the left side of the concentric circles, emotions that are generally generated from reactions occurring in the brain are located. On the right side of the concentric circles, emotions that are generally induced by situational judgment are located. Above and below the concentric circles, emotions that are generally generated from reactions occurring in the brain and induced by situational judgment are located. In addition, the emotion of "pleasure" is located on the upper side of the concentric circles, and the emotion of "displeasure" is located on the lower side. Thus, in the emotion map 400, multiple emotions are mapped based on the structure in which emotions arise, and emotions that are likely to occur simultaneously are mapped close together.

[1696] These emotions are distributed at the 3 o'clock position on the Emotion Map 400, and usually fluctuate between feelings of security and anxiety. In the right half of the Emotion Map 400, situational awareness takes precedence over internal feelings, resulting in a calm impression.

[1697] The inside of the Emotion Map 400 represents inner thoughts, while the outside represents actions. Therefore, the further you go from the outside of the Emotion Map 400, the more visible (expressed in actions) your emotions become.

[1698] Here, human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. Similarly, in robots, cars, motorcycles, etc., emotions can be created based on various balances, such as posture and battery level. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. The emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on a system for analyzing brain physiological signals of speech emotion recognition and emotion, Tokushima University, doctoral dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map contains emotions belonging to a region called "response," where sensation is dominant. The right half of the emotion map contains emotions belonging to a region called "situation," where situational awareness is dominant.

[1699] The emotion map defines two emotions that promote learning. One is the emotion around the middle of the negative "repentance" and "reflection" on the situation side. In other words, it is 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 the emotion around the positive "desire" on the reaction side. In other words, it is when the robot has positive feelings such as "I want more" or "I want to know more."

[1700] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​representing each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple training data sets, which are combinations of user input and emotion values ​​representing each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions located close together have similar values, as shown in the emotion map 900 in Figure 10. Figure 10 shows an example where multiple emotions such as "reassured," "calm," and "confident" have similar emotion values.

[1701] The above description primarily focuses on the functions of the data processing device 12 in relation to this disclosure. However, the system related to this disclosure is not necessarily implemented on a server. The system related to this disclosure may be implemented as a general information processing system. This disclosure may be implemented, for example, as a software program that runs on a personal computer or as an application that runs on a smartphone. The method related to this disclosure may be provided to users in SaaS (Software as a Service) format.

[1702] In the above embodiment, an example was given in which a specific process is performed by a single computer 22. However, the technology of this disclosure is not limited thereto, and a distributed processing of the specific process may be performed by multiple computers, including computer 22. For example, a data generation model 58 may be provided in an external device of the data processing device 12, and the external device may generate data according to the input data.

[1703] In the above embodiment, an example was given in which the specific processing program 56 is stored in the storage 32, but the technology of this disclosure is not limited thereto. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-temporary storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-temporary storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes specific processing according to the specific processing program 56.

[1704] 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.

[1705] Furthermore, it is not necessary to store the entirety 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 the entirety of the specific processing program 56 in the storage 32; it is acceptable to store only a portion of the specific processing program 56.

[1706] The following types of processors can be used as hardware resources to perform specific processing. Examples of processors include a CPU, a general-purpose processor that functions as a hardware resource to perform specific processing by executing software, i.e., a program. Other examples of processors include dedicated electrical circuits, such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices), or ASICs (Application Specific Integrated Circuits), which have circuit configurations specifically designed to perform specific processing. All of these processors have built-in or connected memory, and all of them perform specific processing by using memory.

[1707] The hardware resource that performs a specific process may consist of one of these various processors, or it may consist of 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). Alternatively, the hardware resource that performs a specific process may consist of a single processor.

[1708] Examples of configurations using a single processor include, firstly, a configuration in which one or more CPUs and software are combined to form a single processor, and this processor functions as a hardware resource that performs a specific process. Secondly, there is a configuration using a processor that realizes the functions of the entire system, including multiple hardware resources that perform a specific process, on a single IC chip, as exemplified by SoCs (System-on-a-chip). In this way, a specific process is realized using one or more of the above types of processors as hardware resources.

[1709] Furthermore, the hardware structure of these various processors can more specifically utilize electrical circuits that combine circuit elements such as semiconductor devices. Also, the specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps can be deleted, new steps added, or the processing order rearranged, as long as it does not deviate from the main purpose.

[1710] The descriptions and illustrations presented above are detailed explanations of the technical aspects of this disclosure and are merely examples of the technical aspects. For example, the above descriptions of the structure, function, operation, and effect are examples of the structure, function, operation, and effect of the technical aspects of this disclosure. Therefore, it goes without saying that you may delete unnecessary parts, add new elements, or replace elements in the descriptions and illustrations presented above, as long as you do not deviate from the essence of the technical aspects of this disclosure. Furthermore, in order to avoid confusion and facilitate understanding of the technical aspects of this disclosure, explanations of common technical knowledge and the like that do not require special explanation to enable the implementation of the technical aspects of this disclosure have been omitted from the descriptions and illustrations presented above.

[1711] All documents, patent applications, and technical standards described herein are incorporated by reference to the same extent as if each individual document, patent application, and technical standard were specifically and individually noted to be incorporated by reference.

[1712] The following is further disclosed regarding the embodiments described above.

[1713] (Claim 1)

[1714] A means of receiving audio data and converting it into text data in real time,

[1715] A means for analyzing text data and converting the response generated based on the analysis results into audio data,

[1716] A system including means for transmitting the generated voice data of the response.

[1717] (Claim 2)

[1718] The system according to claim 1, further comprising means for detecting specific keywords from text data and notifying the user.

[1719] (Claim 3)

[1720] The system according to claim 1, further comprising means for enabling a user to intervene in a call when a specific keyword is detected.

[1721] "Example 1"

[1722] (Claim 1)

[1723] A means of receiving audio data and converting it into text data in real time,

[1724] A means for analyzing text data and generating an appropriate response using a generative AI model,

[1725] A means for converting the generated response text data into audio data,

[1726] A means for transmitting the generated response audio data,

[1727] A means of detecting specific keywords and notifying the user,

[1728] A means for users to check notifications and intervene in calls,

[1729] A system that includes this.

[1730] (Claim 2)

[1731] The system according to claim 1, further comprising means for constructing prompt sentences to be input to an AI model that generates text data.

[1732] (Claim 3)

[1733] The system according to claim 1, which allows a user to intervene in a call when a specific keyword is detected.

[1734] "Application Example 1"

[1735] (Claim 1)

[1736] A means of receiving audio data and converting it into text data in real time,

[1737] A means for analyzing text data, generating a response using a generative AI model based on the analysis results, and converting the response into audio data,

[1738] A means for transmitting the generated response audio data,

[1739] A means of obtaining a request from a user as voice data and sending a prompt message to a generating AI model based on that request,

[1740] A means of suggesting appropriate content to the user based on the generated prompt,

[1741] A system that includes this.

[1742] (Claim 2)

[1743] The system according to claim 1, further comprising means for detecting specific keywords from text data and notifying the user.

[1744] (Claim 3)

[1745] The system according to claim 1, further comprising means for enabling a user to intervene in a call when a specific keyword is detected.

[1746] "Example 2 of combining an emotion engine"

[1747] (Claim 1)

[1748] A means of detecting an incoming call and transmitting voice data in real time when the call begins,

[1749] A means for receiving audio data and converting it into text data in real time using recognition means,

[1750] A means for passing text data to an analysis means and creating a response based on the analysis results using a generation means,

[1751] A means for passing the generated response text data to a synthesis means and converting it into audio data,

[1752] A system including means for transmitting generated audio data.

[1753] (Claim 2)

[1754] A means for passing text data and audio data to an emotion recognition means to analyze the user's emotions,

[1755] The system according to claim 1, further comprising means for adjusting the content of the response based on emotion.

[1756] (Claim 3)

[1757] The system according to claim 1, further comprising means for detecting specific keywords from text data, notifying the user when specific emotions and keywords are detected, and enabling the user to intervene in a call.

[1758] "Application example 2 when combining with an emotional engine"

[1759] (Claim 1)

[1760] A means of receiving audio data and converting it into text data in real time,

[1761] A means for analyzing text data and generating a response using a generative AI model based on the analysis results,

[1762] A means for converting the generated response into audio data,

[1763] A means for analyzing a user's emotions from text and audio data using an emotion engine,

[1764] A means for transmitting the generated response audio data,

[1765] A system that includes means of notifying the user when specific emotions and keywords are detected and providing them with options for intervention.

[1766] (Claim 2)

[1767] The system according to claim 1, further comprising means for detecting specific keywords from text data and adjusting the generated response based on a prompt statement.

[1768] (Claim 3)

[1769] The system according to claim 1, further comprising means for enabling a user to intervene in a call when a specific keyword or emotion is detected. [Explanation of symbols]

[1770] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Devices 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robots< / url:> < / url:> < / url:> < / url:>

Claims

1. A means of receiving audio data and converting it into text data in real time, A means for analyzing text data and converting the response generated based on the analysis results into audio data, A system including means for transmitting the generated voice data of the response.

2. The system according to claim 1, further comprising means for detecting specific keywords from text data and notifying the user.

3. The system according to claim 1, further comprising means for enabling a user to intervene in a call when a specific keyword is detected.

Images