Information processing system

Abstract

The application provides an information processing system. An information processing system, characterized by comprising: a processor; wherein the processor is configured to: perform initial processing on an incoming call, extract caller information from a received incoming call signal; perform initial processing on a communication from an unknown caller, compare the caller information with a database to perform risk assessment; use a generative artificial intelligence model to analyze relevant information of the caller and generate prompt information for judging the safety of the caller; use voice recognition technology to analyze the user's request and generate prompt information for initiating communication by artificial intelligence to perform vehicle arrangement and reservation processing; identify user emotions and generate prompt information for providing corresponding services based on the user emotion information.

Description

Technical Field

[0001] The technology disclosed herein relates to an information processing system. Background Technology

[0002] Japanese Patent Application Publication No. 2022-180282 discloses a method for controlling a role-based chatbot executed by at least one processor. The method includes the following steps: receiving a user's speech; adding the user's speech to a prompt word, the prompt word containing instruction statements associated with an explanation of the chatbot's role; encoding the prompt word; and inputting the encoded prompt word into a language model to generate a chatbot response to the user's speech.

[0003] In existing communication technologies, fixed-line telephones and their peripheral systems suffer from several shortcomings in terms of security and intelligent services. Firstly, when receiving incoming calls, terminals typically rely solely on simple caller ID display, failing to effectively assess and identify unknown callers. This makes users vulnerable to telemarketing, harassment, and fraudulent calls. Especially in scenarios where vulnerable groups like the elderly widely use fixed-line telephones, the lack of intelligent caller security assessment mechanisms significantly increases the risk of being scammed. Secondly, traditional fixed-line telephone systems offer limited service functions, typically only providing basic voice call capabilities. They cannot automatically handle complex business processes such as transportation arrangements, reservations, and orders based on users' natural language commands. Users must manually search for numbers, dial, and communicate with service providers, resulting in cumbersome and inefficient operations. Furthermore, existing systems generally lack the ability to recognize and respond to users' emotional states, failing to adaptively adjust service content and interaction methods based on changes in user mood, leading to a poor user experience. Moreover, current terminal devices in optical communication environments primarily perform simple signal conversion and forwarding, failing to fully integrate generative artificial intelligence models and speech recognition technology for unified intelligent management and control of incoming and outgoing calls. Therefore, the main problem to be solved by the present invention is to provide a system that can perform intelligent initial processing and security assessment of incoming calls, automatically generate communication control prompts based on speech recognition and generative artificial intelligence models and execute various appointment and ordering operations, and recognize user emotions and provide services that are more tailored to their needs. The system can also be easily coordinated with optical communication units, thereby significantly improving the security, convenience and intelligent service level of fixed-line telephones and related communication environments. Summary of the Invention

[0004] To address the aforementioned issues, this invention provides an information processing system comprising a processor that executes preset program instructions to achieve comprehensive intelligent control over incoming calls, outbound calls, and user interaction processes. Specifically, the processor is first configured to perform initial processing on incoming calls: upon receiving an incoming call signal from a communication network, the processor automatically extracts caller information from the signal, including the caller's number, call time, and identifier data related to the caller. Subsequently, the processor performs risk assessment processing for communications from unknown callers, comparing the extracted caller information with a pre-established database, which may include blacklists, whitelists, historical call records, and information on external risky numbers, thereby calculating the risk level of the incoming call. Based on the risk assessment results, the processor uses a generative artificial intelligence model to further analyze the caller's relevant information, automatically generating prompts to determine the caller's security. These prompts can be used to drive the terminal to perform specific operations such as interception, hold, transfer, or reminding the user to answer with caution. On the other hand, the processor utilizes speech recognition technology to parse the user's natural language requests, converting the user's speech uttered through a microphone or other input device into text. It then uses a natural language understanding algorithm to identify the user's intent and key parameters (such as time, location, number of people, service type, etc.), and invokes a generative artificial intelligence model to generate corresponding communication control prompts. This enables the AI ​​to automatically initiate communication and complete tasks such as arranging transportation, making restaurant reservations, booking facilities, and ordering various goods or services. Simultaneously, the processor is also configured to analyze the user's voice characteristics, word patterns, and interaction content to identify the user's current emotional state, such as tension, anxiety, anger, or calmness. Based on this emotional information, it generates adaptive service prompts to adjust the interaction strategy with the user. For example, when user anxiety is detected, it provides more safety tips or simplifies operation steps, thereby improving the user experience. Furthermore, the processor is also used to control an extension device connected to the optical communication unit. This extension device converts the optical signal from the optical communication unit into an electrical signal and connects to the communication network. The processor manages the operating status and signal routing of the extension device by generating corresponding prompts and control commands, achieving seamless integration with the optical communication environment. Through the above-described structure and functional configuration, the system of the present invention can realize call security assessment, intelligent outbound call service, emotion perception service and optical communication interface management on a single platform, thereby effectively solving the problems of insufficient security, low level of intelligence and lack of emotion adaptive service in existing technologies.

[0005] "System" refers to an integrated device or combination of devices including at least one processor and connected memory, communication interface and / or other hardware modules, for performing functions such as call processing, voice recognition, risk assessment, generating prompts and controlling communication processes.

[0006] A “processor” is a hardware unit capable of executing program code or instructions to perform operations, logical judgments, and flow control on input data, including but not limited to a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), and a microcontroller or a combination thereof with processing capabilities.

[0007] "Incoming call" refers to a call request or call signal initiated by an external communication terminal or network and transmitted to this system through the communication network. It usually includes the caller ID, signaling information, and optional additional identification data.

[0008] "Incoming call signal" refers to signaling information and / or media data carrying voice used in a communication network to establish, maintain and release voice or data calls, and to indicate that a new call has arrived and to carry information related to the caller.

[0009] "Caller information" refers to a set of information related to the party making the call, including at least one caller number, and may also include the caller's name, operator, geographical location, historical call record identifiers, and other identifying data related to the caller.

[0010] "Unknown caller" refers to a caller whose information has not been registered in the whitelist, contact list, or trusted number list by the system in advance, and is not marked as a known safe or trusted caller in the current database.

[0011] A "database" refers to a storage system used to store blacklists, whitelists, historical call records, risky number information, user preference data, and other structured or unstructured data related to the caller or business processing. It can be a local database, a distributed database, or a database linked to an external server.

[0012] "Risk assessment" refers to the process by which the system compares and analyzes the caller's information with the data stored in the database, and combines preset rules, statistical features and / or machine learning models to calculate and determine the probability that the incoming call may belong to harassment, sales, fraud or other unsafe categories, and gives the corresponding risk level or decision.

[0013] "Generative AI models" refer to AI models that can automatically generate text, prompts, or other content based on input data. These include language models or multimodal models built using technologies such as deep learning and neural networks, used to generate prompts related to communication control, caller analysis, or service execution.

[0014] "Prompt information" refers to text information, control instructions, or semantic structures generated by the system to guide or control subsequent processing steps, including prompt information for determining the caller's security, prompt information for initiating appointments and orders, and prompt information for controlling the working status of communication equipment.

[0015] "Speech recognition technology" refers to the technical process of converting user-input speech signals into corresponding text information. It typically includes steps such as acoustic feature extraction, acoustic modeling, language modeling, and decoding, and is used to enable the system to understand the user's spoken commands or requests in natural language.

[0016] "User request" refers to a user's request to the system via voice or other input methods, which is an operation or service request that the system expects the system to perform, including but not limited to instructions such as calling a car, booking a restaurant, reserving facilities, placing an order, and querying information.

[0017] "AI-initiated communication" refers to the process by which the system, without requiring manual dialing by the user, automatically generates and sends call signals or data requests through the processor-controlled communication interface or terminal to establish contact with external service providers or communication objects to complete corresponding business processes.

[0018] "Transportation arrangement" refers to booking or calling transportation services for users to travel via telephone or other communication methods, including but not limited to the invocation and confirmation of taxis, ride-hailing services, chauffeur services or other transportation services.

[0019] "Reservation" refers to the act of a system contacting a service provider through a communication network on behalf of or assisting a user in reserving service resources at a specific time, place, and under specific conditions, including restaurant reservations, hospital registration, and facility usage reservations.

[0020] "Ordering" refers to the process by which the system, based on user requests, automatically initiates communication to issue instructions for purchasing goods or services in order to complete the order placement, confirmation, and recording. This includes food delivery orders, product purchases, and various on-demand service orders.

[0021] "User emotion" refers to the user's current psychological or emotional state inferred by the system through analysis of information such as the user's voice characteristics, word choice, speech rate, and tone of voice. Examples of such states include tension, anxiety, anger, pleasure, or calmness.

[0022] "Optical communication unit" refers to a device or module used for data or voice communication with external communication networks through optical fiber lines, including optical access terminal equipment that provides optical signal transmission and reception, protocol conversion and network access functions.

[0023] "Apparatus extended to optical communication unit" refers to hardware devices connected to optical communication unit to supplement or extend the functions of optical communication unit, including circuit modules or terminal equipment that perform photoelectric signal conversion, interface adaptation and protocol processing.

[0024] "Optical signal" refers to an optical electromagnetic wave signal that propagates in optical fiber or optical communication medium and is used to carry communication data or control information.

[0025] "Electrical signal" refers to a signal in voltage, current or electromagnetic form that is transmitted in a wire or circuit and is used to represent and carry communication data or control information.

[0026] "Communication network" refers to wired or wireless network infrastructure that enables the transmission of voice, data, or multimedia information, including but not limited to the public switched telephone network, the Internet, cellular mobile communication networks, enterprise private networks, and fiber optic access networks. Attached Figure Description

[0027] Figure 1 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the first embodiment.

[0028] Figure 2 This is a conceptual diagram illustrating an example of the main functions of the data processing apparatus and smart device according to the first embodiment.

[0029] Figure 3 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the second embodiment.

[0030] Figure 4 This is a conceptual diagram illustrating an example of the main functions of the data processing device and smart glasses according to the second embodiment.

[0031] Figure 5 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the third embodiment.

[0032] Figure 6 This is a conceptual diagram illustrating an example of the main functions of the data processing device and head-mounted terminal according to the third embodiment.

[0033] Figure 7 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the fourth embodiment.

[0034] Figure 8This is a conceptual diagram illustrating an example of the main functions of the data processing device and robot according to the fourth embodiment.

[0035] Figure 9 This represents an emotion map that maps multiple emotions.

[0036] Figure 10 This represents an emotion map that maps multiple emotions.

[0037] Figure 11 This is a sequence diagram illustrating the processing flow of the data processing system of the first embodiment.

[0038] Figure 12 This is a sequence diagram illustrating the processing flow of the data processing system in Application Example 1.

[0039] Figure 13 This is a sequence diagram illustrating the processing flow of the data processing system of the second embodiment.

[0040] Figure 14 This is a sequence diagram illustrating the processing flow of the data processing system in Application Example 2. Detailed Implementation

[0041] Hereinafter, an example of an implementation of the system according to the present disclosure will be described with reference to the accompanying drawings.

[0042] First, let me explain the terminology used in the following instructions.

[0043] In the following embodiments, the processor (hereinafter referred to as "processor") with reference numerals may be a single computing device or a combination of multiple computing devices. Furthermore, the processor may be a single computing device or a combination of multiple computing devices. Examples of computing devices include CPU (Central Processing Unit), GPU (Graphics Processing Unit), GPGPU (General-Purpose computing on Graphics Processing Units), APU (Accelerated Processing Unit), etc.

[0044] In the following embodiments, RAM (Random Access Memory), as indicated in the figures, is a memory that temporarily stores information and is used as working memory by the processor.

[0045] In the following embodiments, the memory, as indicated by the reference numerals, is one or more non-volatile storage devices that store various programs and parameters. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), disks (e.g., hard disks), or magnetic tapes.

[0046] In the following embodiments, the communication I / F (Interface) with reference numerals is an interface that includes a communication processor and an antenna, etc. The communication I / F is responsible for communication between multiple computers. As an example of a communication specification applicable to the communication I / F, wireless communication specifications such as 5G (5th Generation Mobile Communication System), Wi-Fi (wireless fidelity) (registered trademark), or Bluetooth (registered trademark) can be listed.

[0047] 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 can be only A, only B, or a combination of A and B. Furthermore, in this specification, when "and / or" connects to express more than three items, the same interpretation as "A and / or B" applies.

[0048] First Implementation Method Figure 1 An example of the configuration of the data processing system 10 according to the first embodiment is shown.

[0049] like Figure 1 As shown, the data processing system 10 includes a data processing device 12 and an intelligent device 14. A server can be cited as an example of the data processing device 12.

[0050] The data processing apparatus 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" as understood in this disclosure. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0051] The smart device 14 includes a computer 36, a receiving device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. In addition, the receiving device 38, output device 40, camera 42, and communication I / F 44 are also connected to the bus 52.

[0052] The receiving device 38 includes a touchscreen 38A and a microphone 38B, and receives user input. The touchscreen 38A receives user input via touch by detecting contact with an indicator (e.g., a pen or finger). The microphone 38B receives user input via sound by detecting the user's voice. The control unit 46A in the processor 46 sends data representing the user input received by the touchscreen 38A and microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data representing the user input.

[0053] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting data in a form perceptible to the user 20 (e.g., sound and / or text). The display 40A displays visual information such as text and images according to instructions from the processor 46. The speaker 40B outputs sound 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 imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0054] Communication I / F44 is connected to network 54. Communication I / F44 and 26 are responsible for sending and receiving various information between processor 46 and processor 28 via network 54.

[0055] Figure 2 The diagram shows an example of the main functions of the data processing device 12 and the smart device 14.

[0056] like Figure 2 As shown, in the data processing apparatus 12, specific processing is performed by the processor 28. A specific processing program 56 is stored in the memory 32. The specific processing program 56 is an example of a "program" as understood in this disclosure. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0057] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. In the emotion inference function (emotion-specific function) using the emotion-specific model 59, various inferences and predictions related to the user's emotions are performed, including inferences and predictions of the user's emotions, but this is not limited to this example. Furthermore, emotion inference and prediction may also include, for example, emotion analysis (parsing).

[0058] In the smart device 14, the processor 46 performs the acceptance output processing. The memory 50 stores the acceptance output program 60. The acceptance output program 60 is used in conjunction with the data processing system 10 and the specific processing program 56. The processor 46 reads the acceptance output program 60 from the memory 50 and executes the read acceptance output program 60 on the RAM 48. The acceptance output processing is implemented by the processor 46 acting as the control unit 46A according to the acceptance output program 60 executed on the RAM 48. Furthermore, the smart device 14 has the same data generation model and emotion-specific model as the data generation model 58 and the emotion-specific model 59, and these models can also be used to perform the same processing as the specific processing unit 290. The acceptance output processing is implemented by the processor 46 acting as the control unit 46A according to the acceptance output program 60 executed on the RAM 48.

[0059] Alternatively, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device (e.g., a generation server) may have the data generation model 58. In this case, the data processing device 12 obtains the processing results (prediction results, etc.) using the data generation model 58 by communicating with the server device that has the data generation model 58. Furthermore, the data processing device 12 may be a server device or a user-held terminal device (e.g., a mobile phone, robot, home appliance, etc.). Next, an example of the processing of the data processing system 10 of the first embodiment will be described.

[0060] Example 1 The flow of a specific process in Example 1 will be described. Each part of the system described below is implemented by the data processing device 12 and the smart device 14. Furthermore, the data processing device 12 is referred to as the "server," and the smart device 14 is referred to as the "terminal."

[0061] In fixed-line communication environments, end users frequently receive calls from unknown senders, including numerous telemarketing, nuisance, and fraudulent calls. Traditional call filtering technologies rely heavily on static blacklists, simple rules, or local pattern matching, which suffer from the following problems: First, blacklist updates are delayed and coverage is limited, failing to effectively address fraudulent activities involving frequently changing numbers; second, fixed-rule filtering methods struggle to comprehensively assess complex call patterns and context, leading to the false blocking of legitimate calls or the allowing of high-risk calls; third, existing systems typically only provide static blocking at the network side, lacking fine-grained control through user interaction, and failing to establish a dynamic decision-making mechanism between "security assessment results" and "user's subjective intentions"; fourth, existing systems rarely use structured feedback from users and historical call data as training data for models, making it difficult to form an adaptive risk assessment mechanism, and hindering the continuous improvement of the system's identification capabilities over time.

[0062] Furthermore, from a computer technology perspective, traditional call handling processes typically involve simply checking if a number exists in a database or making a few conditional judgments in logical branches, failing to fully utilize the ability of generative artificial intelligence models to process unstructured information. At the same time, how to effectively combine number information, historical behavioral data, and prompts from generative artificial intelligence models, and achieve end-to-end security assessment and connection control processes under the constraints of real-time call processing, are also technical challenges that have not yet been properly addressed in existing technologies.

[0063] Therefore, there is a need for a computer implementation scheme that can: automatically identify unknown senders when receiving incoming calls; use generative artificial intelligence models to conduct flexible and high-dimensional security assessments of senders; combine the assessment results with user terminal interaction to achieve refined answer / rejection control; and feed user feedback and assessment results back into learning data to continuously improve the system's identification capabilities and overall security, thereby improving the technical effectiveness of communication systems in terms of security filtering, real-time decision-making, and self-learning capabilities.

[0064] The specific processing performed by the specific processing unit 290 of the data processing apparatus 12 in Embodiment 1 is achieved by the following means.

[0065] In this invention, the server includes: a device for receiving incoming call signals and obtaining and parsing sender identification information from a communication device; a device for comparing the sender identification information with an information set in an information storage device and classifying the sender as a known sender or an unknown sender accordingly; a device for generating a prompt statement containing sender identification information and its associated historical information for unknown senders, inputting the prompt statement into a generative artificial intelligence model to obtain security-related evaluation information, and determining communication security based on the evaluation information; and a device for controlling the connection permission or denial of the communication device based on the security determination result. The system includes: a device for recording the judgment result and sender identification information; a device for sending a notification message containing sender identification information, evaluation information, and security judgment result to an information processing terminal when the security or acceptable status is determined, thereby prompting a user response operation; a device for receiving a connection allow or connection deny response message from the information processing terminal based on the user response operation and controlling the communication device to execute a voice communication connection or interruption based on this and the security judgment result; and a device for storing user response information and security evaluation information as learning data in an information storage device to improve the accuracy of subsequent security judgments for unknown senders. This allows for the construction of a complete, self-learning call security assessment and control link at the computer level, tightly integrating number comparison, generative artificial intelligence model inference, user terminal interaction, and feedback data accumulation. This enables high-precision, low-latency filtering and dynamic decision-making for unknown calls, reducing false positives and false negatives, and continuously optimizing the system's risk identification capabilities over time, thereby significantly improving the computer technology performance related to communication security.

[0066] "Incoming call signal" refers to an electrical or optical signal that is transmitted from a communication network via a communication device and received by the system to indicate the presence of a call request and contains call control information.

[0067] "Communication device" refers to a hardware device used to connect to a communication network and perform call access, call control and media transmission, including but not limited to fixed communication terminals, gateway devices or switching devices.

[0068] "Sender identification information" refers to identification data used to uniquely or quasi-uniquely identify the caller initiator in a communication network, including but not limited to telephone numbers, user identification numbers, or other identifiers associated with the sender.

[0069] "Information storage device" refers to data storage resources used to store information such as sender identification information, historical communication records, security assessment results, and learning data, including but not limited to database systems, file storage systems, or other non-volatile storage media.

[0070] "Information set" refers to a collection of multiple records stored in an information storage device for comparing and classifying sender identification information, including but not limited to a set of trusted numbers, a blacklist set, or a set of historical call records.

[0071] "Known sender" refers to a sender whose sender identification information has been pre-registered or already exists in the information set, and whose security can be determined based on historical records or preset rules.

[0072] "Unknown sender" refers to a sender whose sender identification information is not pre-registered in the information set and lacks clear historical records or preset classification information, requiring further evaluation to determine its security.

[0073] "Prompt statements" refer to natural language or structured text that is automatically generated by the server based on the sender's identification information and its associated historical information, and is provided as input to the generative artificial intelligence model to prompt the model to perform security assessments or related reasoning tasks.

[0074] "Generative AI models" refer to AI models trained through machine learning methods that can generate text or structured outputs based on input prompts. These models are used to perform reasoning and generative processing, including risk assessment and causal explanation generation.

[0075] "Evaluation information" refers to the result data output by the generative artificial intelligence model based on prompt statements, which is related to the sender or the security of its communication, including but not limited to risk level, risk type, reasoning, or other parameters used for security determination.

[0076] "Security determination result" refers to the comprehensive judgment made by the server based on evaluation information and necessary rules or policies regarding whether communication from a specific sender is secure, acceptable, or should be rejected.

[0077] "Information processing terminal" refers to a terminal device used to receive notification information sent by a server and present security results to the user, accept user response operations and return response information to the server, including but not limited to mobile terminals, computing terminals or other electronic devices with display and input functions.

[0078] "Notification information" refers to data sent by the server to the information processing terminal, which includes sender identification information, evaluation information, and security judgment results, and is used to inform users of incoming call status and risk information on the terminal.

[0079] "User response operation" refers to the interactive behavior of users on the interface of the information processing terminal in response to notification information, including but not limited to choosing to answer, reject, mark as spam, or make other feedback inputs.

[0080] "Response information" refers to the control information generated by the information processing terminal based on the user's response operation and sent to the server, which is used to instruct the execution of connection allow, connection rejection or other processing for a specific incoming call.

[0081] "Voice communication connection" refers to the communication channel established between the sender and the communication device for transmitting voice media data when an incoming call is allowed to be connected.

[0082] "Voice communication interruption" refers to the process of releasing or terminating the voice media channel and related signaling session by controlling the communication device when an incoming call is rejected or the call ends.

[0083] "Risk differentiation" refers to the result of classifying communication risks into several levels or categories based on evaluation information, which is used to indicate the possible security level of incoming calls, including but not limited to low risk, medium risk, and high risk.

[0084] "Summary description" refers to explanatory text or summary information generated by a generative artificial intelligence model or server to briefly explain the reasons for risk differentiation.

[0085] "Visual display method" refers to the display strategy by which information processing terminals present risk differentiation and notification information on the user interface in the form of different colors, icons, text styles or layouts.

[0086] "Optical signal conversion device" refers to a hardware device used to perform bidirectional conversion between optical signals and electrical signals, thereby enabling connection with optical fiber communication lines and acquiring incoming voice signals or sending control signals.

[0087] "Learning data" refers to a collection of data consisting of user response information, security evaluation information, and related contextual information, which is stored for training, fine-tuning, or optimizing generative artificial intelligence models or related decision-making algorithms.

[0088] "Subsequent security assessment and processing" refers to the process by which the server, based on existing learning data, updated models, or rules, reassesses and determines the security of a new call received from the same or similar unknown sender.

[0089] In one embodiment of the invention, the server connects to a fixed communication device, an information processing terminal, and a generative artificial intelligence model runtime environment via a communication network. The server's hardware may consist of a general-purpose computing device, including, for example, a multi-core central processing unit, main memory, non-volatile memory, a network interface, and a graphics processing unit or dedicated accelerator for running the generative artificial intelligence model. The server's software may run a general-purpose operating system, such as a UNIX-like operating system, and deploy a database management system (such as a relational database management system), application server middleware, and HTTP / HTTPS communication components on it.

[0090] In this embodiment, the program executed by the server can be stored on a non-volatile storage medium and loaded into main memory, and executed sequentially or in parallel on the central processing unit. When executing the program, the server performs multi-stage processing on data related to incoming call signals, including: parsing communication signaling, structuring data for sender identification information, retrieving and matching data records in information storage devices, converting the format of input and output data for generative artificial intelligence models, and generating and issuing control commands.

[0091] In the connection section with the fixed communication device, the server controls the optical signal conversion device through a communication interface. For example, the optical signal conversion device converts optical signals from the fiber optic line into Ethernet electrical signals, and the server receives the corresponding packet data through the network interface. The server uses the communication protocol stack module to parse signaling messages, extracting fields such as the calling number, the called number, and the session identifier from the signaling messages, and encapsulating these fields into internal data structures, such as record structures or key-value mappings. The server writes the sender identification information into the call session table of the database and establishes a unique session identifier for each incoming call.

[0092] When determining whether a sender is unknown, the server uses a database query module to retrieve information from multiple sets. The server can maintain at least three tables in the database: a trusted number table, a blacklist table, and a historical call table. The server uses sender identification information as an index key to access these tables, quickly locating the existence of a corresponding record through an index structure (such as a B+ tree index or a hash index), thus completing the classification process internally with low time complexity. Because the server uses structured storage and indexing techniques, it can maintain low-latency classification even in high-concurrency call scenarios, which is beneficial for meeting strict time constraints in real-time call control scenarios.

[0093] After identifying an unknown sender, the server generates a prompt message to provide to the generative AI model. The server maintains a set of template rules in its prompt message generation module. The template can contain a fixed explanatory section and replaceable variable sections, such as phone number, historical call count, and user feedback tags. The server reads historical records related to the number from the database, including past risk levels, whether the user answered the call, and whether it was marked as a spam call, and embeds this information into the prompt message in natural language.

[0094] For example, the server can generate the following prompt: "Please analyze whether the following phone number is safe: Phone number: 03-1234-5678. Please refer to typical patterns of fraudulent calls, characteristics of telemarketing harassment calls, publicly available blacklist information, and historical records (e.g., 2 out of the last 3 calls were rejected by the user, and 1 was marked as suspected telemarketing), and give a risk level (low / medium / high) and a brief reason." For example, the server can generate the following prompt when more granular differentiation is needed: "Please conduct a risk assessment for phone number 03-9876-5432. Known history: This number has received 5 calls in this system, of which 3 were marked as harassment and 2 were marked as suspected fraud. Please assign a risk level (low / medium / high) to this number based on common telecom fraud patterns, and explain the main basis for your judgment in Chinese no more than 100 characters." After generating the prompt statement, the server packages the prompt statement and related metadata (such as language type and application scenario identifier) ​​into an internal data object, and converts it into a message format suitable for network transmission through the serialization module. Then, it sends it to the generative artificial intelligence model server through the HTTP / HTTPS client component.

[0095] The server can use a neural network-based text generation model in the generative AI model module. The server can deploy this model on a separate computing node or call it through an external service interface. Structurally, the generative AI model can adopt a multi-layered self-attention neural network architecture, such as an encoder-decoder or decoder stack structure. The model can include several embedding layers, multi-head self-attention layers, feedforward network layers, and normalization layers. The input is a prompt statement that has been segmented into words or sub-words, and the output is text containing the risk level and reason.

[0096] The server can use supervised learning methods during the model training phase. After collecting a large amount of historical data with risk labels and user feedback, the server combines the prompt statements with the corresponding target output (e.g., "Risk level: High, main reason: repeatedly marked as a scam call") into training samples. During training, the server defines cross-entropy loss as the objective function and updates the weights in the network through the error backpropagation algorithm. The server can use stochastic gradient descent or its variants (such as adaptive moment estimation) for weight updates and can improve the robustness of the model through data augmentation methods (e.g., paraphrasing the prompt statements, randomly truncating parts of the context, etc.).

[0097] When the server receives the model output during the inference phase, it typically receives a piece of natural language text. The server parses this text using a parsing module, extracting structured information such as "risk level" and "reasoning points" from the text through keyword matching, regular expressions, or lightweight classification models. For example, if the model output text contains the phrase "risk level: high," the server sets the risk level field to "high"; if the output contains a fragment stating "marked as a scam call on multiple public blacklists," the server stores that fragment as a reasoning field in the database.

[0098] After completing the security assessment, the server binds the results to the session state and stores them. The server records not only the final risk level, but also the model name, model version number, warning message content, output text, and parsing results. The server stores these fields in a structured manner in the database, which is beneficial for subsequent analysis of model performance, retraining, or reviewing the basis for specific assessments. Through this approach, the server improves upon the "black box judgment" problem in traditional systems, making the security assessment process auditable and interpretable within the computer.

[0099] When the server confirms that an incoming call is selectable by the user, it sends a notification message to the terminal. When constructing the notification message, the server adds a display level based on the risk level, such as "Safe," "Medium Risk," or "High Risk." The server can set different identifiers for different levels, allowing the terminal to display different colors or icons. The notification message includes sender identification information, risk level, a summary of the reason, the time of the incoming call, and a session identifier.

[0100] In this embodiment, the terminal can be a mobile terminal or a fixed-location information processing terminal. The terminal includes a processor, memory, display device, and input device (such as a touchscreen or buttons) in its hardware. The terminal runs a terminal application in its software, receives notification information from the server, and displays it on the user interface. After parsing the notification information, the terminal adopts different interface layouts according to the risk level; for example, a red warning bar and prominent icon are used for high-risk calls, a yellow prompt is used for medium-risk calls, and a normal style is used for low-risk calls. The terminal displays the phone number, a risk level description, and a brief explanation in the interface, such as: "Risk Level: High. Explanation: This number is marked as a suspected impersonation of a financial institution on multiple public blacklists." While displaying the interface, the terminal provides interactive options such as answering, rejecting, and marking as spam calls. After the user selects an option via touch or button operation, the terminal converts this action into a response message and sends it to the server. The terminal can locally cache user operation data for a short period, such as operation delay time and whether the user requested detailed reasons. This data, after being uploaded to the server, helps the server further analyze user behavior patterns, thereby adjusting prompts or threshold strategies.

[0101] In this embodiment, users do not need to understand the internal details of the generative artificial intelligence model; they only need to make simple selections based on the prompts on the terminal interface. Unlike traditional systems where users make subjective judgments about unfamiliar numbers, in this invention, user decisions are based on complex data analysis and model reasoning performed by the server.

[0102] After receiving the response information from the terminal, the server performs control based on a combination of the security assessment result and the user's selection. For example, when the risk level is "low" and the user selects "answer," the server sends an connection command to the fixed communication device, controlling the optical signal conversion device and the communication device to establish a voice media channel. When the risk level is "high" and the user does not actively select to answer or explicitly selects "reject," the server sends a disconnect command to the fixed communication device, terminating the call session on the network side. During the control phase, the server reduces unnecessary media channel establishment, lowers the communication load on the network side, and improves the overall system throughput by releasing resources in a timely manner.

[0103] When the server stores user feedback and model output together as learning data, it generates a complete data record for each incoming call. This record includes: the prompt message content, the model output text, the parsed risk level, the user's action (answer / reject / mark as spam), the actual call result (e.g., call duration, whether there was an abnormal hang-up), and whether the user subsequently modified the marker. During subsequent offline training or online fine-tuning, the server extracts samples from these records. It can calculate the loss function using the consistency between the model's risk level and the user's actual feedback as a metric and reduce errors by updating the model parameters. As data accumulates, the server can achieve higher accuracy on the same hardware resources, thereby reducing false positives and false negatives.

[0104] Through the aforementioned mechanisms, the server not only automates the human judgment process but also improves the caller ID process at the computer technology level. On one hand, the server enhances the speed of number matching and historical retrieval through structured storage and efficient indexing. On the other hand, the combination of prompt templates and generative artificial intelligence models enables security assessments to utilize contextual and historical behavioral information for high-dimensional reasoning—something difficult to achieve with traditional fixed-rule or simple blacklist mechanisms. Furthermore, the server continuously feeds user feedback back into model training data, forming a closed-loop adaptive evolution mechanism that improves the system's judgment capabilities over time. This demonstrates a significant technological advantage compared to systems that rely solely on manual, periodic rule maintenance.

[0105] Servers can employ different types of generative AI models in various implementation variations. For example, in one variation, the server can deploy a smaller model in a local computing environment to reduce network latency and enhance privacy. In another variation, the server can employ a multi-model collaborative approach, such as using a lightweight classification model to initially screen numbers, and only calling a complex, large-scale generative AI model for a subset of high-uncertainty or high-risk candidate numbers, thereby reducing overall computational load and improving response speed.

[0106] The server can also adjust the risk level threshold according to application requirements. For example, in environments with extremely high security requirements, the server can default to classifying "medium-risk" numbers as "suspicious" and tend to reject them; while in environments with higher requirements for service connection rates, the server can present detailed reasons for "medium-risk" numbers to the terminal and let the user decide whether to answer. This method of adjusting the judgment behavior through parameterized strategies enables the present invention to adapt to the technical requirements of different scenarios.

[0107] Through the above implementation, the server establishes a dynamic risk assessment mechanism within the computer, centered on prompt statements and generative artificial intelligence models. By controlling communication devices and optical signal conversion devices, it transforms abstract text reasoning results into specific control operations on real communication links, realizing an integrated technical link of "data processing - model reasoning - device control". This has resulted in significant technical effects in the field of call security filtering, including improved security assessment accuracy, reduced processing latency, increased resource utilization, and adaptive optimization of model performance during long-term operation.

[0108] use Figure 11 The processing flow is explained.

[0109] Step 1: The server receives incoming call signals through the communication device and parses the sender's identification information.

[0110] Input: Signaling data from the communication network (including calling number, called number, signaling type, etc.).

[0111] The server performs protocol parsing on the input signaling data. Based on the field format of the preset communication protocol (such as SIP or vendor proprietary protocol), the server extracts fields from the message header and message body, parses the field representing the telephone number into sender identification information, and parses the session identifier into an internal session ID.

[0112] Output: Structured incoming call record data, which includes at least the session ID, sender identification information, called number, and reception time.

[0113] Step 2: The server retrieves the sender identification information from the information storage device and determines whether the sender is an unknown sender.

[0114] Input: Sender identification information and session ID output from step 1.

[0115] The server uses sender identification information as the query key to perform an index search in the trusted number set, blacklist set, and historical call set. The server compares the database index structure (such as a B+ tree or hash index) to see if there is an exact match. The server processes the data based on the search results: if a match exists in any set, the status is marked as "known sender"; otherwise, it is marked as "unknown sender".

[0116] Output: Session state data with sender category labels (known / unknown), and a summary of possible historical records (e.g., number of historical calls, historical risk level).

[0117] Step 3: The server generates prompts for generative artificial intelligence models when the sender is unknown.

[0118] Input: The sender identification information marked as "unknown sender" and its history summary (if it exists).

[0119] The server fills in placeholders in a predefined prompt template with information such as phone number, historical call count, and past risk tags. The server then performs string concatenation and formatting operations to generate complete natural language text.

[0120] The server generates the following prompt statement for example: "Please analyze whether the following phone number is safe: Phone number: 03-1234-5678. Please refer to typical patterns of fraudulent calls, characteristics of telemarketing harassment calls, publicly available blacklist information, and historical records (e.g., 2 out of the last 3 calls were rejected by the user, and 1 was marked as suspected telemarketing), and give a risk level (low / medium / high) and a brief reason." Output: The prompt text associated with the current session ID, and the input data object used to invoke the generative artificial intelligence model.

[0121] Step 4: The server invokes a generative artificial intelligence model to perform a risk assessment on the prompt statements.

[0122] Input: The prompt text generated in step 3 and the sender identification information.

[0123] The server encodes the prompt and additional metadata into a request message via an HTTP / HTTPS client component and sends it to the generative AI model service interface. Internally, the generative AI model performs operations such as vectorization, attention calculation, and feedforward network computation to extract features and infer from the input text, generating output text that includes a risk level and a reasoning explanation.

[0124] The server receives the response message from the model service, decodes it, and obtains the original output text.

[0125] Output: The model output text corresponding to the current session, such as "Risk level: High" or "Reason: Repeatedly marked as fraud".

[0126] Step 5: The server parses the output of the generative artificial intelligence model and structures the evaluation information.

[0127] Input: The model output text obtained in step 4.

[0128] The server performs text analysis on the output text, extracting fields such as "risk level" and "reasoning" from the natural language using keyword matching, regular expressions, or lightweight parsing algorithms. The server maps the risk level to an internal enumeration value (e.g., 0 = low, 1 = medium, 2 = high) and saves the reasoning as a string field. The server generates a structured evaluation record, associated with the session ID and sender identification information.

[0129] Output: A data structure containing evaluation information fields such as risk level, explanation of reason, and model identifier.

[0130] Step 6: The server generates a security assessment result based on the evaluation information.

[0131] Input: The evaluation information output from step 5 and the system's preset risk threshold rules.

[0132] The server makes logical judgments based on risk level values ​​and preset policies. For example, when the risk level is "low," the security judgment result is set to "safe"; when the risk level is "medium," the result is set to "suspicious"; and when the risk level is "high," the result is set to "dangerous." The server can also modify the initial judgment result by combining historical statistics (such as the number being marked as spam by users multiple times). The server writes the final judgment result into the session log.

[0133] Output: Security assessment result (safe / suspicious / dangerous) associated with the current session and the updated session state.

[0134] Step 7: The server controls the connection behavior of communication devices based on the security assessment results.

[0135] Input: Security assessment result and session ID.

[0136] The server processes the input results in two branches: if the result is "dangerous," the server generates a connection rejection control command (e.g., hang-up or rejection signaling) and sends it to the fixed communication device via the communication interface, triggering a call interruption; if the result is "safe" or "suspicious," the server keeps the current session resources in a pending state, waiting for further feedback from the terminal and user. During this process, the server updates the session resource table, such as marking the session as "awaiting user decision" or "rejected."

[0137] Output: The result of sending control commands to the communication device or optical signal conversion device, and the updated session resource status.

[0138] Step 8: The server sends a notification containing security information to the terminal.

[0139] Input: Security assessment results, evaluation information (risk level and reason), sender identification information, and session ID.

[0140] The server encapsulates these input fields into a notification message structure, adds display suggestions based on the risk level (such as "use red warning icon"), and sends the notification message to the terminal via push service or real-time connection protocol.

[0141] Output: Terminal-oriented notification message, which includes the phone number, risk level, reason summary, timestamp, and session ID.

[0142] Step 9: The terminal receives the notification and displays call details and risk information to the user.

[0143] Input: The notification message sent by the server.

[0144] The terminal parses the received message, extracting the phone number, risk level, reason text, and interface display parameters. The terminal then draws interface elements on the display screen, showing the number and risk information, using different colors, icons, or layouts according to the risk level. Simultaneously, the terminal generates clickable controls on the interface such as "Answer," "Reject," and "Mark as Spam Call."

[0145] Output: The call details screen displayed on the user interface, and the interface status awaiting user action.

[0146] Step 10: Users respond based on the information displayed on the terminal.

[0147] Input: The phone number, risk level warning, and reason displayed on the terminal interface.

[0148] Users read the risk warnings and explanatory text on the interface, and choose the appropriate action based on their own judgment, such as clicking the "Answer" or "Reject" button, or selecting "Mark as Spam Call". The user's physical actions are captured by the terminal input module and converted into internal events.

[0149] Output: User response event, indicating the user's selection type (answer / reject / mark as spam).

[0150] Step 11: The terminal converts the user's response into response information and sends it to the server.

[0151] Input: User response event and session ID from the notification message.

[0152] The terminal constructs a response information data structure based on the operation type, filling in the session ID and operation type, such as "accept", "reject", or "mark_spam". The terminal sends this response information to the server via HTTP / HTTPS or a real-time communication channel and updates the display status locally (e.g., displaying "connecting" or "rejected").

[0153] Output: The response message transmitted to the server, and the updated interface status on the terminal.

[0154] Step 12: The server ultimately controls the voice communication connection or interruption based on the user's response information.

[0155] Input: The response information sent by the terminal, the session ID, and the existing security assessment results.

[0156] The server parses the response information to obtain the user's selection. The server combines the user's selection with the security judgment result: when the judgment is "secure" and the user selects "answer", the server sends an connection command to the communication device to control the establishment of a voice media channel; when the judgment is "suspicious" and the user selects "answer", the server still allows the connection but records the user's tolerance behavior in the database; when the user selects "deny" or "mark as spam", the server sends a disconnect command, releases session resources, and updates the status of the number in the blacklist set or marking table.

[0157] Output: The final control action (connect or disconnect) performed on the communication device, as well as the update results of session records and number tag status.

[0158] Step 13: The server stores the evaluation information and user feedback from this session as learning data.

[0159] Input: Evaluation information (risk level and reason), user response information, call result (whether the call was connected, call duration, etc.).

[0160] The server combines these input fields into a single learning data record and writes it to the learning data table. The server standardizes some fields, for example, mapping risk levels to numerical labels, encoding user actions as categorical variables, and dividing call durations into intervals. During subsequent model training or fine-tuning, the server extracts samples from this learning data table to update the parameters of the generative AI model.

[0161] Output: The updated learning data storage state, and dataset entries that can be used for subsequent training.

[0162] Step 14: The server periodically optimizes the generative artificial intelligence model based on accumulated learning data.

[0163] Input: Multiple sample records accumulated in the learning data table.

[0164] During the offline training phase, the server loads these samples, using the prompt statements as input and the target output (e.g., "true risk label" or "user-confirmed labeling result") as the supervision signal. It calculates the deviation between the model output and the target using an error function (e.g., cross-entropy loss). The server iteratively updates the weights of each layer in the model using backpropagation and adjusts the parameters using optimization algorithms (e.g., adaptive learning rate methods). After training, the server deploys the updated model weights to the inference environment, making subsequent evaluations of unknown senders more accurate.

[0165] Output: Updated generative AI model parameters and a performance-enhanced model version for subsequent online inference.

[0166] Application Example 1 The process flow corresponding to the specific processing in Use Case 1 will be described below. The various parts of the system described below are implemented by the data processing device 12 and the intelligent device 14. Furthermore, the data processing device 12 is referred to as the "server" and the intelligent device 14 is referred to as the "terminal".

[0167] In existing communication technologies, risk control for unknown incoming calls typically relies on static blacklists, simple rules, or single scoring models, which presents the following problems: First, servers only block calls based on whether the number appears in the list, failing to dynamically assess multiple dimensions such as call frequency, time period, historical call duration, and user feedback. This results in slow response to newly emerging fraudulent numbers and high false positive and false negative rates. Second, there is a lack of fine-grained collaborative control mechanisms between servers and terminals. Servers cannot issue refined processing strategies (such as automatic blocking, warning only, or call reservation and forwarding) based on real-time risk assessment results, and terminals struggle to promptly present easily understandable risk explanations to users, thus impacting user decision-making efficiency and security experience. Third, existing systems mostly use fixed... The system relies on rules or traditional models for judgment, lacking the ability to perform interpretable analysis of communication risks using generative artificial intelligence models and prompting mechanisms. It cannot flexibly generate explanatory texts and processing recommendations based on different numbers and scenarios. Fourth, the large amount of feedback data generated by users during call answering, rejecting, and marking processes is not systematically collected and structured for use. The server cannot automatically convert this feedback into training data, making it difficult for the model to continuously adapt, update, and improve recognition accuracy. Fifth, in voice interaction scenarios, the control logic for automatic appointments, order placement, and proxy outbound calls is separated from the security assessment logic. It cannot unify the arrangement of voice recognition results, generative artificial intelligence model outputs, and external communication control within the same system, increasing system complexity and reducing reliability.

[0168] Therefore, a computer implementation scheme is needed that can achieve multi-source communication feature fusion, prompt-driven generative artificial intelligence model invocation, automatic risk assessment and strategy decision-making, terminal collaborative display and control, and closed-loop self-learning based on user feedback on the server side. This would improve the ability to identify unknown incoming calls, the accuracy of security interception, and the overall intelligence of communication services from the perspectives of system structure and data processing flow, thereby improving the computer technology itself.

[0169] The specific processing performed by the specific processing unit 290 of the data processing apparatus 12 in Application Example 1 is achieved by the following means.

[0170] In this invention, the server includes a processing unit for receiving incoming call signals and extracting communication identification information; a judgment unit for retrieving and determining unknown incoming calls from the information storage unit; a feature generation unit for generating statistical feature information based on historical communication records and user evaluations; a model calling unit for constructing and outputting model input containing prompt statements and structured features, calling a generative artificial intelligence model to perform risk assessment and obtain risk level indicators and explanatory information; a strategy decision unit for automatically generating processing strategies such as communication blocking, warning display, call hold transfer, or normal connection by combining threshold information and user setting information; a terminal coordination control unit for sending risk information and processing instructions to the terminal device and controlling the terminal display and call connection status; a feedback acquisition unit for collecting user answering operations and number marking information on the terminal and storing them in a structured manner as training data; and a model updating unit for periodically extracting call history information and user evaluation information to train or fine-tune the generative artificial intelligence model to continuously improve the accuracy of unknown incoming call risk assessment. This allows for the formation of a dynamic risk assessment and control mechanism on the server side, with prompt statements as the entry point, generative artificial intelligence models as the core, and user feedback as the closed loop. This mechanism not only enables refined risk identification and multi-level processing strategy distribution for unknown incoming calls, but also improves the automation and intelligence of communication security processing from both the system architecture and algorithm process levels through structured feature generation and continuous model updates. This, in turn, enhances the data processing capabilities and overall performance of computer systems in communication security scenarios.

[0171] "Incoming call signal" refers to signaling information transmitted by the communication network and received by the communication device, used to indicate that a call request has arrived and the related control information it carries.

[0172] "Communication identification information" refers to identification data used to identify one or more parties in a communication, including but not limited to the caller's number, the called number, the caller's name, and other identification parameters associated with the communication session.

[0173] "Information storage unit" refers to a data storage device used to store user-related information, call history information, user evaluation information, model parameters and threshold information, etc., and may consist of one or more storage media and their management programs.

[0174] "User-related information" refers to configuration information, contact information, trust lists, block lists, and other data associated with a specific user or user terminal, used to determine communication relationship attributes.

[0175] "Unknown sender" refers to a communication status where the sender's communication identification information is not registered in the user's relevant information as a known or trusted communication initiator.

[0176] "Known sender" refers to a communication status where the communication identification information has been registered in the user's relevant information as belonging to a known or trusted communication initiator.

[0177] "Historical communication record information" refers to a collection of data generated and recorded in past communication sessions, related to the number of calls, call duration, call direction, call time, and processing results.

[0178] "User evaluation information" refers to the subjective marking or classification results made by users on specific communication identification information, including but not limited to label data that marks numbers as fraud, sales, normal, service, etc.

[0179] "Statistical feature information" refers to numerical or categorical features generated through statistical processing based on historical communication records and user evaluation information to describe the characteristics of communication behavior, including communication frequency information, call duration information, time period information, and user feedback information.

[0180] "Structured information" refers to a set of characteristic data that is organized and represented according to a predetermined data structure, suitable for direct processing by computer programs and models, including characteristic data represented in the form of fields, key values, or tables.

[0181] "Prompt statements" refer to natural language or semi-structured text content constructed for interaction with generative artificial intelligence models. This text contains task descriptions, input conditions, output format requirements, and information related to communication identification.

[0182] "Generative artificial intelligence models" refer to artificial intelligence models that can generate text, numerical values ​​or other forms of output results based on input information. They use parameterized functions to reason about input data and output results including risk indicators, risk descriptions or communication content.

[0183] "Risk index" refers to a numerical or graded indicator used to represent the degree of risk of a certain communication, such as fraud, harassment, or unsolicited sales pitches.

[0184] "Risk level explanation information" refers to explanatory information that explains the causes or basis of risk level indicators in natural language or readable text form, based on the output of generative artificial intelligence models.

[0185] "Threshold information" refers to one or more numerical boundaries that are pre-set and stored to distinguish different risk levels or trigger different processing strategies.

[0186] "User settings information" refers to configuration information that is pre-set and stored by the user to specify the processing strategies that the system should take under different risk levels.

[0187] "Processing strategy" refers to the processing method performed in the system for a specific communication, including communication blocking, warning display, call hold and normal connection.

[0188] "Communication blocking" refers to the process of controlling the establishment of a communication connection to prevent a specific incoming call from establishing a voice or data channel with the terminal.

[0189] "Warning display" refers to the act of presenting text, icons, or color codes that indicate communication risks to users on the display of a terminal device in order to alert them.

[0190] "Holding the call" refers to a process where, after a communication request arrives, a final call connection with the user's terminal is not immediately established. Instead, the call is held on the network side and then transferred to the terminal only after the user confirms the request.

[0191] "Normal connection" refers to the process of establishing a communication connection according to the standard procedure without imposing additional restrictions on incoming calls.

[0192] "Terminal device" refers to an electronic device used by a user to receive information sent by a server and to perform display, input, and call control, including portable terminals, fixed terminals, and other devices with communication functions.

[0193] "Display unit" refers to a display component installed on a terminal device or other output device, used to present text, graphics, icons, or color information to the user.

[0194] "Call connection control unit" refers to the hardware or software module used to control the establishment, maintenance and release of communication sessions, and to execute the process of allowing or denying connection according to control instructions.

[0195] "Connection operation information" refers to operation record data generated by users answering, rejecting, hanging up, or other call control-related operations on their terminal devices.

[0196] "Number classification information" refers to the categorization results made by users on communication identification information, including the label data that classifies the identification information as fraud, sales, normal or other categories.

[0197] "Training data" refers to the set of sample data used to train or fine-tune the parameters of a generative artificial intelligence model, which typically includes input features and corresponding labels or expected outputs.

[0198] "Data processing" refers to the process of cleaning, transforming, normalizing, extracting features, and formatting raw data to facilitate subsequent statistical processing and model training.

[0199] "Statistical processing" refers to the process of summarizing and analyzing data through counting, summing, averaging, calculating proportions, or other statistical methods.

[0200] "Model parameters" refer to the adjustable values ​​or weights used within a generative artificial intelligence model to determine the mapping relationship between inputs and outputs.

[0201] "Training or fine-tuning" refers to the process of updating model parameters by executing learning algorithms on training data to improve model performance. Fine-tuning refers to a form of training that involves making small adjustments to parameters based on an existing model.

[0202] "Assessment processing" refers to the process of using generative artificial intelligence models to reason about specific communication identification information and its related characteristics in order to output risk indicators and explanatory information.

[0203] "Speech recognition processing" refers to the signal processing and pattern recognition process that converts a user's speech signal into corresponding text or natural language representation.

[0204] "Natural language information" refers to user intent or content information obtained through speech recognition or other methods and expressed in the form of natural language text.

[0205] "Appointment processing information" refers to structured or semi-structured data generated based on user natural language information and used to execute appointment-related services (such as appointment time, location, service content, etc.).

[0206] "Order processing information" refers to structured or semi-structured data generated based on user natural language information and used to execute order placement or purchase-related business (such as product type, quantity, price, delivery information, etc.).

[0207] "Communication content generation and processing" refers to the process of using generative artificial intelligence models to generate text content or instruction content for external communication based on prompts and input data.

[0208] "External communication network" refers to a public communication network, private communication network, or data network outside the system, used for voice or data communication with external entities.

[0209] "Outbound communication" refers to the communication connection process initiated by the system to a target identifier in an external communication network.

[0210] "Control information" refers to instruction data used to instruct communication devices, terminal devices, or network entities to perform specific control actions (such as establishing a connection, refusing a connection, switching paths, etc.).

[0211] "Signal conversion communication device" refers to a communication device that can perform bidirectional or unidirectional conversion between optical signals and electrical signals to realize signal transmission between different physical media.

[0212] The "communication path control unit" refers to the functional module used to control the switching and routing of signals between different transmission paths (including optical signal paths and electrical signal paths).

[0213] "Connection mode" refers to the combination of connection methods, path selection, bandwidth configuration, and related connection parameters between a communication device and a communication network.

[0214] "Optical signal path" refers to the physical or logical channel through which communication signals are transmitted via optical fiber or other optical transmission media.

[0215] "Electrical signal path" refers to the physical or logical channel through which communication signals are transmitted via cables, lines or other electrical transmission media.

[0216] In one embodiment of the present invention, the server, terminal, and user each assume different functional roles. Through the collaboration of hardware devices and software modules, intelligent risk assessment and communication control for unknown incoming calls are achieved. The system structure, data structure, generative artificial intelligence model structure, learning method, specific prompting statements, and technical effects are described in detail below.

[0217] I. System Overall Structure In this embodiment, the server is configured as one or more computer devices, including at least one central processing unit (CPU), an optional graphics processing unit (GPU), main memory, non-volatile memory, a network interface, and an operating system. In one example, the server may run a general-purpose server operating system and install database management software (e.g., relational database management software), a data processing framework (e.g., a library for performing data statistics), and generative artificial intelligence model inference services on it.

[0218] In this embodiment, the terminal can be a smartphone terminal, a tablet computer terminal, or a desktop computer terminal. The terminal includes a display unit, an input unit (touchscreen, buttons), a local storage unit, a communication interface, and a local communication control interface. The terminal runs a terminal application program, which maintains a network connection with the server to receive risk information, display prompts, and collect user feedback.

[0219] In this implementation, users can view call risk warnings, answer or reject calls, and mark numbers as feedback via their terminals.

[0220] II. Server-side functional modules and data structures In this embodiment, the server includes multiple functional modules, each of which can be physically executed by the same processor and is logically divided into the following units.

[0221] 1. Signaling Reception and Identification Information Extraction Module The server includes a signaling receiving module. It receives incoming call signals from the communication network via a communication interface, including call setup control signaling. The server uses a signaling parsing component to extract communication identification information from the signaling, such as the calling number, called number, call identifier, and operating area information. The server creates a session record structure in main memory for each incoming call. This structure includes at least a session ID field, a communication identification information field, a timestamp field, and a subsequent risk assessment result field.

[0222] 2. Information storage department and data table structure The server creates various data tables in the information storage department to support subsequent feature generation and model training. Exemplary table structures include: The server stores user identifiers and user settings (such as automatic blocking thresholds and alert policies) in the "User Information Table".

[0223] The server stores the user's registered phone number and its category in the "Contacts Table" to distinguish between known and unknown senders.

[0224] The server stores call history information in the "Call Record Table," including the number, call direction, call time, call duration, processing result (answered, rejected, not connected), and the user to which the call belongs.

[0225] The server stores user-assigned tag information for phone numbers in the "User Evaluation Table," including the phone number, tag category (fraud, sales, legitimate, service, etc.), tag time, and user identifier.

[0226] The server stores threshold information, model version number, model parameter path information, and feature configuration in the "model configuration table".

[0227] The server stores training sample indexes extracted and transformed from real-time data in the "training sample table" for subsequent batch training or fine-tuning processes.

[0228] 3. Unknown Sender Detection Module After receiving the communication identification information, the server queries the contact table and user-related information. If the corresponding number does not exist in the user's contacts or trusted list, the server marks the number as an unknown sender in the session record structure. The server uses an index query during this process to reduce retrieval time and writes the determination result into the session record.

[0229] 4. Feature Generation Module and Feature Structure In the feature generation module, the server reads historical records related to the current number from the call log table and the user rating table. The server constructs statistical feature information, which consists of multi-dimensional fields; examples include: The server generates numerical characteristics such as "number of incoming calls in the past 30 days", "number of rejected calls in the past 30 days", "average call duration in the past 30 days", "percentage of calls received at night", and "percentage of calls received on weekdays".

[0230] The server generates counting features such as "number of times marked as fraud", "number of times marked as sales pitch", and "number of times marked as normal" based on the user rating form.

[0231] The server can map the time field to discrete features of "time period categories" (such as 0-6 o'clock, 6-12 o'clock, 12-18 o'clock, 18-24 o'clock) to form a time pattern code.

[0232] The server performs data normalization processing, such as taking the logarithm of the number of calls, scaling the call duration, and standardizing the scale value, in order to form a feature vector with a uniform range.

[0233] The server organizes the above fields into structured information, which can be implemented as a multidimensional vector or a set of key-value pairs. The server stores this structured information together with the current communication identification information in the session record.

[0234] 5. Prompt Statement Construction Module The server constructs a prompt statement before invoking the generative artificial intelligence model. The server uses templated text and dynamic interpolation to insert communication identification information and statistical feature information into the Chinese prompt text. The prompt statement generated by the server in one example is as follows: In one example, the server uses the following prompt statement: "You are a phone number risk assessment assistant. Based on the number information and historical statistics provided below, please determine whether this number is likely to be fraudulent or used for telemarketing. Please give a risk score from 0 to 100 and briefly explain your reasoning."

[0235] Telephone number: +819012345678.

[0236] Number of calls received in the past 30 days: 12, 8 of which were rejected by users. The average call duration was 10 seconds. Most of the calls occurred on weekday mornings. Many users marked the number as 'telemarketing'.

[0237] Please return the results in JSON format, with the following fields: risk_score (integer), risk_level (low / medium / high), and reason (Chinese explanation). In another example, the server uses the following message for numbers with no historical records: "Please act as a phone number security analysis assistant and analyze the following information to determine if the phone number may be a scam or a sales call."

[0238] Telephone number: +8613800000000.

[0239] Historical data: No historical records; this is the first time this message has appeared in this system. The current call time is 10 PM on a weekday.

[0240] Please provide a decimal risk score between 0 and 1, and explain the reason in one sentence in Chinese. Output format: score: number; reason: explanation. The server explicitly transmits the task objective, input conditions, and output format constraints to the generative artificial intelligence model through the aforementioned prompts, thereby forming a controllable reasoning path within the model.

[0241] III. Generative Artificial Intelligence Model Structure and Learning Methods 1. Model Structure In this embodiment, the server employs a large-scale generative artificial intelligence model based on deep neural networks. The model is generally an encoder-decoder structure or an autoregressive structure, which includes a multi-layer attention mechanism module, a multi-head attention sublayer, a feedforward fully connected sublayer, and a normalization sublayer.

[0242] In one example, the server concatenates the prompt and structured features into text input. The model first maps the input characters or words into vector representations through an embedding layer. The server uses a multi-layer self-attention mechanism in the model's encoding section to model the relevance of different parts of the prompt.

[0243] The server generates a text-based risk score and explanation during the model decoding process. The server can explicitly include key markers such as "risk_score," "risk_level," and "reason" in the output to make post-processing parsing more stable.

[0244] In another embodiment, the server inputs structured features as additional numerical vectors into a dedicated feature channel of the model. The server maps the numerical features to vectors of the same dimension as the text embeddings through a feedforward network and fuses them at a specific layer of the model, enabling the model to make judgments using both textual context and statistical features.

[0245] 2. Learning and Fine-tuning Methods During the model training phase, the server extracts training samples from the information storage department. The server constructs each sample into a format of "prompt statement + label data". The label data includes risk level, risk score, category label, etc., which are determined manually or based on user evaluation.

[0246] The server uses either cross-entropy loss or mean squared error loss during training. If the model is designed to generate text, the server uses a language modeling loss function and performs maximum likelihood training on the target output text sequence.

[0247] In some implementations, the server employs supervised fine-tuning: based on a general language model, the server fine-tunes it using call records and user-labeled data collected by the system. The server updates model weights using batch gradient descent, performing forward propagation, calculating loss, backpropagation, and updating parameters in each training batch.

[0248] To improve generalization ability, the server performs data expansion operations during the training phase, such as randomly rewriting the expressions in the prompts with synonyms and abstracting the time descriptions, thereby enhancing the model's adaptability to different forms of expression.

[0249] Through the aforementioned structure and learning method, the server enables generative artificial intelligence models to perform fine-grained assessments of communication risks and generate readable outputs containing explanatory information, thus avoiding becoming an unexplainable "black box."

[0250] IV. Algorithmic Characteristics and Technical Effectiveness of Risk Assessment and Strategy Decision-Making In this invention, the server does not simply perform simple database queries and fixed rule judgments, but combines multi-source feature fusion with generative artificial intelligence model reasoning to form an unconventional processing flow.

[0251] In the strategy decision-making unit, the server first obtains the risk level indicators and explanatory information output by the model, and then combines "threshold information" and "user settings information" to perform multi-level decisions. The server uses multiple threshold ranges and logical operations, for example: The server sets high-risk and medium-risk thresholds, and maps "high-risk + automatic blocking enabled" to "direct blocking" based on whether the user has enabled automatic blocking; maps "medium-risk + warning only mode" to "warning displayed but connection allowed"; and maps "low-risk and good historical interaction" to "reserved transfer or normal connection".

[0252] Through the aforementioned decision-making logic, the server maps complex model outputs into explicit communication control commands, thereby directly controlling the call connection on the terminal side.

[0253] Compared to traditional solutions based solely on static blacklists, the server in this implementation can: By combining statistical features with generative artificial intelligence models, the server can quickly provide a risk assessment for newly appearing numbers, thereby improving the detection coverage of new types of fraudulent numbers.

[0254] By continuously integrating user feedback data and fine-tuning the model, the server enables the risk assessment model to gradually approach real user perception during actual operation, thereby reducing the false alarm rate.

[0255] The server encodes complex, multi-dimensional information into an input space that the model can understand by generating unified structured features and constructing prompts, thereby improving the consistency and interpretability of the model output.

[0256] The aforementioned improvements are technically reflected in enhanced processing accuracy, reduced error recognition rate, shortened communication control decision time, and optimized server-side resource utilization efficiency.

[0257] V. Terminal-side collaborative display and control In this embodiment, the terminal has a client application installed.

[0258] After receiving the risk level information and processing instructions from the server, the terminal parses the information locally. The terminal maps the "risk level index," "risk grade," and "reason explanation" to interface elements, such as text prompts, color markings, and icon displays.

[0259] The terminal calls the local communication control interface on the terminal side to perform specific operations according to the server's policies, such as blocking incoming calls, not displaying the ringing interface, and only displaying pop-up warnings.

[0260] After the user completes a call, the terminal displays a feedback interface prompting the user to select options such as "Mark as Scam," "Mark as Telemarketing," or "Mark as Normal." The terminal uses the user's selection as number classification information, packages it together with the call duration and processing result, and uploads it to the server.

[0261] In this way, the terminal not only serves as a display and input interface, but also plays the role of the front end for data collection, enabling the server to obtain high-quality training samples.

[0262] VI. Voice Interaction and Outbound Call Control Scenarios In a certain implementation, the server extends this system for use in scenarios involving automatic appointments or automatic order placement.

[0263] In this scenario, the server receives voice data from the terminal. The terminal first converts the voice into natural language text using a local or cloud-based speech recognition module and sends it to the server. The server then extracts the business intent and parameters from the natural language text, such as "Reserve a vehicle for 10:00 AM tomorrow" or "Deliver goods to a certain address."

[0264] The server generates appointment processing information or order processing information based on the extracted results. Then, the server constructs corresponding prompt statements, embeds the processing information and context descriptions into the text, and sends it to the generative artificial intelligence model, asking it to generate natural language content suitable for external communication, such as explanatory language for service providers.

[0265] After receiving the generated communication content, the server uses it as an outbound communication script to control the communication device to initiate a call to the external communication network and drive the speech synthesis or interactive process according to the generated text.

[0266] During this process, the server can simultaneously reuse the aforementioned risk assessment mechanism to conduct security assessments on the target number or service endpoint, preventing the abuse of automated outbound calls.

[0267] VII. Extended Forms of Signal Path Control In another embodiment, the server further controls the communication device capable of converting between optical and electrical signals.

[0268] Based on the risk assessment results and user settings, the server generates different path control instructions. For example, when processing high-priority trusted calls, it prioritizes the low-latency optical signal path, while when processing general calls, it selects the conventional electrical signal path.

[0269] The server switches between optical and electrical signal paths by controlling the communication path control unit, reducing latency and jitter in critical business calls, thereby achieving a balance between communication quality and resource utilization.

[0270] VIII. Summary of Technical Effects and Causal Relationships Through the collaboration of the aforementioned multiple modules, the server enables the system to improve upon computer technology itself in the following aspects: The server upgrades traditional static rule filtering into a dynamic, self-learning risk assessment system through specific data structures (multi-table databases, structured feature vectors) and specific processing flows (feature generation, prompt statement construction, generative artificial intelligence model reasoning, and policy decision-making), thereby improving the accuracy and robustness of the processing.

[0271] By designing prompt templates and feature fusion methods for generative AI models, the server enables the model output to include not only probabilistic scores but also explanatory text. This helps achieve more stable parsing and decision-making in post-processing and reduces the impact of model output uncertainty on system behavior.

[0272] The server continuously collects user feedback from the terminal side, converts it into training data in real time, and performs regular fine-tuning on the generative artificial intelligence model. This allows the model parameters to continuously adapt to the latest fraud methods and communication patterns, significantly reducing the workload and time delay of manually maintaining the blacklist.

[0273] By directly mapping risk assessment results into call connection control commands and path control commands, the server enables the invention to go beyond abstract judgment and directly affect the working state of communication devices, thereby achieving specific control over real communication equipment and producing practical technical effects in areas such as communication security and resource scheduling.

[0274] By using local caching and a fast judgment mechanism, the terminal can provide preliminary prompts before the server returns the complete evaluation results, reducing the impact of network latency on user experience and security, and reducing the average response time and communication load from the perspective of the overall communication system.

[0275] In summary, this invention, through collaboration among the server, terminal, and user, utilizes a generative artificial intelligence model and prompting mechanism to form a complete technical solution across three levels: data structure, algorithm flow, and device control. This transforms the identification and handling of unknown call risks from a simple manual task into an automated process, representing a substantial improvement in computer and communication technologies in terms of model structure, feature utilization, data closed-loop, and communication control.

[0276] use Figure 12 The processing flow is explained.

[0277] Step 1: The server receives incoming call signals and extracts communication identification information.

[0278] The server's input is incoming signaling data from the communication network (including raw signaling fields such as caller ID, called number, and call identifier). The server performs protocol parsing and field separation on the signaling data, resolving fields such as caller ID, called number, call time, and call ID from binary or text format into internal structure fields. The server's output is a session object containing communication identification information, including at least the normalized caller ID, called number, and timestamp.

[0279] Step 2: The server queries the information storage department based on the communication identification information to determine whether the incoming call is from an unknown sender.

[0280] The server's input consists of the caller ID and user identifier from the session object output in step 1. The server performs index queries on this number in the contact table, trust list, and blacklist, comparing the number string or its hash value to see if a record exists. The server performs string matching and Boolean checks, writing the results to the session object's status field. The server's output is an updated session object marked with either "Unknown sender" or "Known sender".

[0281] Step 3: The server generates statistical feature information for numbers that are determined to be unknown senders.

[0282] The server's input consists of the caller ID and user identifier from a session object marked as an unknown sender. The server queries the call log table by number to retrieve historical records, performing statistical operations such as counting, summing, and averaging on the query results to calculate the number of incoming calls, rejected calls, average call duration, and call ratios for different time periods over a recent period. Simultaneously, it retrieves data from the user rating table to count the number of times the number was marked as fraudulent, sales-related, or legitimate. The server normalizes, partitions, or performs logarithmic transformations on these values, combining them into a structured feature vector or key-value pair set. The server's output is a set of statistical feature information associated with the number, which is written into the feature fields of the session object.

[0283] Step 4: The server constructs prompt statements based on communication identification information and statistical feature information.

[0284] The server's input consists of the caller ID from the session object and the statistical feature information generated in step 3. Based on a predefined text template, the server inserts fields such as caller ID, number of calls, rejection rate, average call duration, time period distribution, and user tagging into placeholder positions, and then generates a complete Chinese prompt statement through string concatenation. The server's output is a prompt statement text containing a task description, input conditions, and output format requirements, which is used for subsequent calls to the generative artificial intelligence model.

[0285] Step 5: The server inputs prompts and structured features into the generative artificial intelligence model to perform risk assessment reasoning.

[0286] The server's input consists of the prompt text generated in step 4 and the structured feature information from step 3. The server takes the prompt text as input, encodes the structured features as additional input (e.g., concatenating them into a text description or converting them into numerical vectors), and invokes the deployed generative AI model's inference service via a network request. Internally, the model segments or partitions the input text, maps words to embedding vectors, performs vector operations through multiple attention mechanisms and feedforward networks, and generates text containing risk scores and risk descriptions at the output. The server's output is the risk assessment result text returned by the generative AI model, which includes numerical values ​​representing the risk level indicators and risk description information.

[0287] Step 6: The server parses the output of the generative artificial intelligence model and generates a numerical risk index.

[0288] The server's input is the risk assessment result text obtained in step 5. The server extracts the specific content of "risk score," "risk level," and "reasoning explanation" from the text through string matching, regular expressions, or simple delimiters. It converts the score field to a numeric type, maps the risk level to an internal enumeration value, and saves the explanation text as a string. The server performs range checks and outlier handling to ensure the values ​​are within preset ranges. The server's output is a risk assessment structure containing numerical risk indicators, discrete risk levels, and explanation text, which is written to the session object and stored in the risk assessment table.

[0289] Step 7: The server generates a processing strategy based on risk indicators, threshold information, and user settings.

[0290] The server's input consists of the risk assessment structure generated in step 6, as well as the threshold information and user settings stored in the information storage unit. The server compares the risk score with high-risk and medium-risk thresholds through numerical comparison calculations to determine the basic risk level. Then, based on whether the user has enabled automatic blocking, warning only, or call reservation / forwarding, it sets execution condition branches to determine the final processing strategy type (e.g., blocking, warning, call reservation / forwarding, normal connection). During this process, the server performs Boolean logic operations and conditional judgments, writing the results to the "Processing Strategy" field of the session object. The server's output is a strategy information structure that clearly identifies the processing method, including the action type, display text, and parameters to be passed to the terminal.

[0291] Step 8: The server sends risk information and processing instructions to the terminal.

[0292] The server's input consists of the processing strategy information generated in step 7 and the risk assessment results from step 6. The server packages fields such as risk score, risk level, cause description, and processing action type into a message object, serializes it, and sends it to the corresponding terminal via a network communication protocol. Before sending, the server can compress and sign the data to reduce bandwidth consumption and ensure integrity. The server's output is a control message indicating successful delivery to the terminal, and it records the sending time and status locally.

[0293] Step 9: The terminal receives messages from the server and updates the caller ID interface.

[0294] The terminal's input consists of risk information and processing instructions from the server. The terminal parses this message in its local communication module, extracting the risk score, risk level, explanatory text, and processing action fields. Based on the risk level, the terminal maps it to different colors, icons, or prompts, overlaying these visual elements onto the system's incoming call interface. If the processing action is blocking, the terminal calls the local call control interface to reject the call; if the processing action is warning, the terminal allows ringing and displays a warning message; if the processing action is hold and forward, the terminal displays a dialog box asking the user to confirm whether to answer. The terminal's output is the updated user interface state and possible local call control actions.

[0295] Step 10: Users can answer or reject calls based on the risk warnings displayed on the terminal.

[0296] User input consists of risk warnings, phone number information, and selectable operation buttons displayed on the terminal interface. Users select actions such as "answer," "reject," or "do not process" via touchscreen, buttons, or other input methods. User actions are recorded on the terminal as connection operation information, including operation type, operation time, and current call status. User output represents the final human decision signal for the incoming call, which is captured by the terminal and used for subsequent control and feedback.

[0297] Step 11: The terminal controls the call connection and collects feedback information based on user operations.

[0298] The terminal's input consists of the user's operation commands in step 10 and the processing strategy issued by the server in step 8. Based on the operation type, the terminal calls the system call control interface to establish or release a voice channel, and then answers or hangs up. After the call ends or is rejected, the terminal displays a feedback interface prompting the user to select a number category (e.g., scam, sales, legitimate). The terminal organizes the user's selected category along with the call number, call duration, and processing result into a feedback data record. The terminal's output is a feedback data structure containing connection operation information and number category information.

[0299] Step 12: The server receives feedback from the terminal and updates call history and user evaluation information.

[0300] The server's input consists of feedback data records uploaded by the terminal, including the phone number, call duration, operation result, user-selected category tag, and timestamp. The server writes this data to the call record table and user rating table, updating the count field for any existing records of the same number. During updates, the server performs insert or update operations and maintains necessary indexes. The server's output is the updated historical data and rating data, used for subsequent feature generation and model training.

[0301] Step 13: The server constructs training samples from accumulated data and periodically trains or fine-tunes the generative artificial intelligence model.

[0302] The server's input consists of call logs and user ratings stored in the information storage unit, as well as the current model configuration. The server extracts samples according to predetermined rules, such as selecting numbers that have been frequently marked recently, combining their statistical features with user tags to construct sample pairs of "input features + target risk level / score". The server performs data cleaning, missing value handling, and encoding transformation on the samples, and then groups them into training batches. During training, the server feeds the constructed prompts and feature information into the model, calculates the error between the model output and the target label, uses a loss function (such as cross-entropy or mean squared error) to obtain the gradient, and updates the model parameters using an optimization algorithm. The server's output is the updated generative AI model version and the new version information recorded in the model configuration table.

[0303] Step 14: The server uses the updated model and the latest threshold configuration for risk assessment and decision-making in subsequent call processing.

[0304] The server's input consists of the new model parameters generated in step 13, the updated threshold configuration, and the communication identification and statistical characteristics of the new incoming calls. During inference, the server invokes the updated generative AI model to execute the same computational process as in steps 5 through 7, using the improved parameters obtained during training to evaluate unknown numbers. The server's output is the risk assessment results and processing strategies generated based on the latest model, thereby providing higher accuracy and faster response risk control capabilities in the next round of call processing.

[0305] Alternatively, an emotion engine for inferring user emotions can be combined. That is, the specific processing unit 290 can also use the emotion-specific model 59 to infer user emotions and perform specific processing using user emotions.

[0306] Example 2 The flow of a specific process in Example 2 will be described. Each part of the system described below is implemented by the data processing device 12 and the smart device 14. The data processing device 12 will be referred to as the "server," and the smart device 14 as the "terminal."

[0307] The technical problem to be solved by this invention belongs to the field of computer technology, and in particular relates to the application and performance improvement of human-computer interaction, natural language processing and network communication control in reservation and ordering scenarios.

[0308] In existing technologies, voice-based appointment or ordering systems typically only recognize speech as text and then process it through rule engines or fixed procedures, which has the following shortcomings in computer technology: (1) Limited natural language understanding ability: Traditional systems rely on keyword matching or simple rules, which cannot accurately extract multiple parameters such as intent, time, location, and number of people in user statements, resulting in the need for multiple rounds of interaction or manual intervention, increasing processing delay and wasting computing resources.

[0309] (2) Lack of a unified prompt statement driving mechanism: In existing systems, the invocation of generative artificial intelligence models (such as large-scale language models) is often static and single-task. The "prompt statements" are not uniformly managed as programmable control objects, making it difficult to dynamically adjust the model behavior according to different business scenarios (ride-hailing, restaurant reservation, security judgment, etc.), resulting in low efficiency of computing resource utilization and poor scalability.

[0310] (3) Separation of communication control and semantic parsing: In the existing architecture, the natural language parsing module and the communication interface control module are often implemented by different subsystems. There is a lack of standardized mapping and automatic generation mechanism between structured semantic information and underlying call control parameters, which leads to complex system design, redundant code, and a lot of manual development required when adding new business types.

[0311] (4) Lack of comprehensive computational processing of user emotions and security: Traditional voice customer service or appointment systems rarely infer the user's emotional state automatically from voice or text, and also lack a general computational framework for risk assessment of unknown callers and security judgment in combination with generative artificial intelligence models. As a result, they cannot dynamically adjust dialogue strategies, service processes and communication permissions at the computer system level, which reduces the overall intelligence and security of the system.

[0312] (5) Insufficient support for multiple communication media: The existing system has limited ability to integrate heterogeneous communication resources such as optical communication devices and different types of network interfaces. It lacks a technical solution that uses generative artificial intelligence models and prompts to uniformly generate control information and automatically extend it to the underlying communication devices, which leads to the need to re-customize a large amount of control logic when the system is extended to new communication media.

[0313] Therefore, how to provide a computer implementation system based on a generative artificial intelligence model and a prompt statement-driven mechanism, so that: ● Complete end-to-end processing from voice acquisition, natural language understanding, structured instruction generation, database query, communication control to result feedback under a single architecture; ● It can dynamically generate appropriate prompts based on user statements and emotional states, guiding the generative AI model to output suitable structured data and control information; ● Capable of performing automated risk assessments on unknown callers and integrating them with communication control strategies; ● It can generate control information that is compatible with different communication interfaces (including electrical interfaces and optical communication devices) in a unified manner; Improving human-computer interaction efficiency, automation of communication control, and system security and scalability at the computer technology level is a technical issue that this invention urgently needs to address.

[0314] The specific processing performed by the specific processing unit 290 of the data processing apparatus 12 in Embodiment 2 is achieved by the following means.

[0315] In this invention, the server includes: a processing unit for receiving digital voice data uploaded by a terminal and converting the digital voice data into text information through a voice recognition processing module; a processing unit for automatically generating a prompt statement for calling a generative artificial intelligence model based on the text information, user identifier, terminal location, and contextual session state, and inputting the text information and the prompt statement into a natural language processing engine containing a generative artificial intelligence model to parse structured instruction information containing user intent, reservation conditions, and control parameters from the text information; a processing unit for accessing an information storage device based on the structured instruction information to retrieve service provider information records to obtain contact information and business capability information; a communication control unit for automatically generating communication control information based on the structured instruction information and the contact information, and initiating a call or data communication through a communication service interface to complete reservation or order processing in transportation services, facility reservations, or other business scenarios; and a communication control unit for constructing a user-oriented result data structure based on the reservation result information returned through the communication service interface or service provider device, and further generating... The system comprises: a result generation unit that drives a generative artificial intelligence model to generate prompts for natural language notifications, thereby obtaining notification messages and sending them to a terminal; an emotion analysis unit that estimates the user's emotional state from the text information and / or the digital voice data, and dynamically adjusts the content of the prompts and the expression of the notification messages input to the generative artificial intelligence model based on the emotional state to optimize the human-computer interaction experience; an emotion analysis unit that receives incoming call signals from external communication networks, extracts caller information, compares the caller information with data in an information storage device, performs rule-based or model-based risk assessments on unknown callers, and generates prompts for security judgments, which are then input into the generative artificial intelligence model to obtain a structured security judgment result reflecting the risk level and processing strategy, and controls whether to establish, restrict, or terminate communication; and a communication extension unit that, when an optical communication device is present, controls the optical communication device to perform optical signal to electrical signal conversion and access the corresponding communication network based on control information output by the generative artificial intelligence model under the guidance of the prompts, thereby achieving extended control of multiple communication media within the same semantic control framework. This allows for the formation of a unified processing framework within the computer system, with prompt statements as the core control mechanism, generative artificial intelligence models as the semantic and policy engine, and structured instruction information as the intermediate representation. This enables end-to-end automated processing from voice input to communication control. While ensuring the accuracy of semantic understanding, it reduces the complexity of hard-coding rules, improves the scalability to new business scenarios and new communication media, and enhances interaction efficiency, security, and resource utilization at the system level by integrating sentiment analysis and call risk assessment. This substantially improves the human-computer interaction and communication control processing methods in the field of computer technology.

[0316] "Voice input" refers to the sound signal generated by the user through vocalization and collected by a voice input / output device for subsequent digital processing and voice recognition.

[0317] "Digital voice data" refers to audio data in digital form, obtained by converting analog voice signals through audio acquisition and analog-to-digital conversion, and expressed at a predetermined sampling rate and quantization precision.

[0318] "Voice input / output device" refers to a hardware device used to acquire a user's voice and output audio information to the user when necessary, including microphones, speakers and their driving circuits, etc.

[0319] "Speech recognition processing" refers to the computational processing of digital speech data, including acoustic feature extraction, acoustic modeling, and language modeling, in order to convert speech content into text information.

[0320] "Text information" refers to user instructions expressed in character or text form, obtained through speech recognition processing or other input methods.

[0321] A "natural language processing engine" refers to a software module or system used for natural language processing operations such as semantic analysis, intent recognition, and parameter extraction of textual information.

[0322] "Generative artificial intelligence models" refer to artificial intelligence models that are based on machine learning algorithms and can automatically generate text, structured data, or control information based on input prompts and text data.

[0323] "Prompt statements" refer to input statements or instruction texts constructed to guide generative artificial intelligence models to perform specific tasks, and are used to specify the model's role, objectives, and output format.

[0324] "Structured instruction information" refers to a data set that is formed based on the results of natural language parsing and represents user intent, reservation conditions, and related control parameters in a predefined field and data structure.

[0325] "Reservation conditions" refer to the parameter information used to limit the service content during the reservation or order processing, including time, location, number of people, service type, etc.

[0326] "Information storage device" refers to a storage system used to persistently store contact information, service provider information, risk assessment information, and log data, including databases, file storage, or other storage media.

[0327] "Contact information" refers to information related to the service provider used to establish communication or conduct business access, including telephone numbers, network addresses, identifiers, etc.

[0328] "Service provider" refers to a business entity or system that provides transportation services, catering services, or other reservation or ordering services to users.

[0329] "Communication service interface" refers to the software and hardware interface used to initiate calls or data communication through a communication network, including application programming interface, communication protocol stack and related driver modules.

[0330] "Communication control information" refers to the set of control parameters used to control the communication service interface to perform call establishment, data session establishment and related parameter configuration.

[0331] "Reservation result information" refers to structured data returned by the service provider or communication service interface that indicates the reservation or order processing result, including status, time, location, confirmation number, etc.

[0332] "Notification message" refers to text or speech content generated based on appointment result information and used to present the processing result to the user.

[0333] "Terminal device" refers to a computing device operated by a user for inputting voice and receiving notification messages, including mobile terminals, fixed terminals, or other devices with audio and network functions.

[0334] "User emotional state" refers to state information inferred from text information and / or digital voice data that indicates the user's emotional tendency or attitude, such as pleasure, anger, tension, etc.

[0335] The “emotion analysis unit” refers to a functional module used to identify and estimate a user’s emotional state from text information or digital voice data.

[0336] "Caller information" refers to the identification information extracted from the incoming call signal to identify the initiator of the communication, including number information, address information or other identifiers.

[0337] "Risk assessment" refers to the calculation process of judging the security and trustworthiness of communication with an unknown caller based on the caller's information and stored data.

[0338] "Security assessment prompts" are prompts containing risk assessment information designed to guide generative artificial intelligence models in outputting security levels or processing strategies.

[0339] "Security control unit" refers to a functional module that controls the establishment, restriction or termination of communication based on the security judgment result.

[0340] "Optical communication device" refers to a communication hardware device used to receive or transmit optical signals and realize photoelectric conversion.

[0341] "Communication extension unit" refers to a functional module used to control optical communication devices or other heterogeneous communication devices to connect with and extend the operation of communication networks based on the control information output by generative artificial intelligence models.

[0342] In the implementation of this invention, the server, terminal, and user each play different functional roles in the system. The server mainly performs computational processing such as speech recognition result processing, generative artificial intelligence model invocation, structured instruction generation, database retrieval, communication interface control, and security assessment; the terminal mainly performs speech acquisition, speech recognition invocation, result display, and local interaction; the user interacts with the terminal via voice. The following describes the system's hardware structure, software module composition, data structure, and algorithm processing methods in different implementation scenarios.

[0343] I. Overall System Composition In one implementation, the server is deployed in a data center or cloud computing platform. The server includes: a general-purpose central processing unit (CPU, such as a multi-core processor), a graphics processing unit (GPU, such as a general-purpose GPU supporting large-scale matrix operations), random access memory (RAM), persistent storage (such as a solid-state drive), a network interface controller (NIC), and optional security modules. The server operating system can be a general-purpose server operating system, such as a UNIX-like system or a server-specific operating system.

[0344] The server includes the following modules at the software level: 1. Voice Recognition Access Module: Used to receive audio data or voice recognition results uploaded by the terminal, and to call external voice recognition APIs (such as general voice recognition service interfaces) or locally deployed voice recognition engines when needed.

[0345] 2. Natural Language Processing Engine: Includes generative artificial intelligence models, rule parsing modules, entity recognition modules, time parsing modules, etc.

[0346] 3. Prompt Statement Generation Module: This module is used to construct different prompt statements based on business scenarios, user context, emotional state, and security assessment results to drive the generative artificial intelligence model.

[0347] 4. Structured Instruction Generation Module: Used to convert natural language or structured text output by generative artificial intelligence models into strictly defined data structures.

[0348] 5. Database Access Module: Used to access relational databases or key-value databases to retrieve service provider information, user configurations, risk records, etc.

[0349] 6. Communication Interface Control Module: Used to interface with various communication service interfaces, including telephone call interfaces, data communication interfaces, and control interfaces for optical communication devices.

[0350] 7. Security Assessment and Incoming Call Risk Control Module: Used to assess the risk of incoming call information and feed the results back to natural language processing and communication control.

[0351] 8. Sentiment Analysis Module: Used to estimate the user's emotional state based on text and / or voice features, and adjust prompts and response strategies accordingly.

[0352] 9. Log and monitoring module: Used to record the running status, request parameters and results of each module for subsequent analysis and optimization.

[0353] In one implementation, the terminal can be a mobile or fixed terminal equipped with a microphone, speaker, display screen, and network connectivity. The terminal runs a general-purpose mobile operating system or an embedded operating system and has client applications installed. The terminal applications include an audio acquisition module, a voice recognition module, a UI display module, a TTS (text-to-speech) module, and a communication module with a server.

[0354] In everyday use, users simply need to speak to the terminal and receive feedback through screen display or voice broadcast, without needing to understand the internal technical details.

[0355] II. Server-side program structure and data structure The server in this invention employs a modular software architecture. The server maintains multiple core data structures in memory to support efficient collaboration between generative artificial intelligence models and communication control. 1. User session context structure The server maintains a session context data structure for each user, containing the user's identifier, text information from recent rounds of dialogue, a list of generated structured instructions, location coordinates, preference settings, and an estimated sentiment state. This structure helps the server incorporate historical information when generating prompts, thereby improving the consistency and accuracy of the generated results.

[0356] 2. Structured instruction information structure The server parses the output of the generative artificial intelligence model into structured instruction information with clearly defined fields. This structure may include: - intent (intent type, such as calling a ride, making a restaurant reservation, canceling a reservation, etc.); - time_slot (normalized timestamp or time range); - location_slot(geographic object, containing latitude and longitude, address text, etc.); - party_size (number of people parameter); - service_type (service type); - preference (price, distance, rating, and other preference parameters); - safety_level (safety level label); - emotion_state (emotional state label).

[0357] The server uses this structure to directly drive database queries and control communication interfaces, avoiding complex text parsing again, thereby improving processing speed and reducing error rate.

[0358] 3. Telephone and Communication Control Information Structure The server maintains the communication control information structure in the communication interface control module, including: - target_endpoint (target number or network address); - call_type (voice call, data session, etc.); - signaling_parameters; - media_parameters (encoding / decoding formats, etc.); - retry_policy (retry policy); - timeout_setting (timeout period).

[0359] By mapping structured instruction information to this control information structure, the server can quickly generate control parameters that are adapted to different communication platforms.

[0360] III. Generative Artificial Intelligence Models and Prompt Statement Mechanisms In one implementation, the server uses a generative AI model based on the Transformer architecture. This model includes multi-layered self-attention encoders and decoders, supporting joint modeling of long text prompts and user input. The server deploys the model on GPUs to accelerate matrix multiplication and attention computation.

[0361] During the model training phase, the server uses a large-scale dialogue corpus, reservation scenario corpus, and security judgment-related corpus for pre-training and fine-tuning. The server employs the following training method: - Loss function: Cross-entropy loss, used to minimize the difference between the model-generated sequence and the target sequence; - Optimization algorithm: Adaptive gradient optimization method; - Learning rate scheduling: warm-up + cosine annealing strategy; - Regularization: Dropout is added to the attention layer and feedforward network to reduce overfitting; - Data augmentation: Add synonym rewriting, order perturbation, and noise injection to the training text to improve the model's robustness to diverse expressions.

[0362] During the inference phase, the server controls the model's behavior through prompts. These prompts are not fixed templates but are dynamically constructed by the prompt generation module based on the business scenario and system state. For example, when parsing a user's ride-hailing request, the server generates the following prompt: "You are a natural language parsing module for a booking and ride-hailing system."

[0363] User input: 'Please call a taxi to my office building.' Please extract the intent and parameters, and output the following: intention: Place: time: Number of people: If a parameter is not mentioned, enter 'Not specified'. When natural language results for user prompts are needed, the server generates different prompt statements, for example: "You are a user-friendly system prompt generation module."

[0364] The system's internal results are as follows: Service type: Taxi reservation Departure time: 18:10 Pick-up location: Near the user's current location License plate number: XXXXXX Please generate a concise and clear Chinese prompt to display to the user. The server uses this prompt-based control method to allow the generative AI model to switch between different subtasks (intent recognition, parameter extraction, result presentation, security instructions, etc.) without having to train a different model for each subtask. This improves the efficiency of computing resource utilization and reduces the overhead of data copying and conversion between modules.

[0365] IV. Characteristics and Technical Effects of Server-Side Algorithms The server adopts a "structured intermediate representation" strategy in its overall algorithm design: 1. Instead of directly using the output of the generative AI model for underlying communication control, the server first converts the output into structured instruction information, and then a specialized mapping algorithm generates communication control information. This introduces a verifiable data layer between the model output and the physical communication device, facilitating error detection and recovery.

[0366] 2. The server uses rule constraints and vocabulary constraints during the structured instruction generation process. For example, the server imposes a finite set constraint on the intent field, requires the time field to conform to a standard time format, and requires the location field to contain coordinates or a resolvable address. Compared to pure natural language output, this approach significantly reduces the probability of abnormal data entering the communication control module, thereby reducing false call rates and resource waste.

[0367] 3. The server combines vector retrieval with traditional keyword retrieval when searching the database. The server can vectorize restaurant or service provider descriptions and calculate similarity with user preference vectors to select more suitable service providers. Simultaneously, the server still uses the index structure for range filtering (e.g., distance, price, opening hours) to achieve efficient and accurate retrieval.

[0368] Through the above design, the server achieves the following effects at the computer technology level: - Due to the unified management of prompt statements and the structured intermediate representation, the server reduces the need for hard-coding rules for different business operations, reduces program complexity, and facilitates maintenance and expansion.

[0369] - Because structured instruction fields are subject to rule constraints, error propagation paths are shortened, communication initiation error rates are reduced, and overall system stability is improved.

[0370] - Because generative AI models are confined to a specified task space, the search space during the inference phase is significantly reduced, the model response time is shortened, and the server's concurrent processing capability is improved.

[0371] V. Specific Implementation of Sentiment Analysis and Safety Assessment The server uses a specialized classification model in its sentiment analysis module. The server can employ a dual-channel architecture: one channel is based on text (textual information), and the other is based on speech features.

[0372] 1. Textual Emotion Channel The server uses a multi-layer bidirectional encoder (such as a bidirectional Transformer encoder) to encode text sequences and adds a classification layer on top to output emotion labels (such as positive, neutral, negative, angry, etc.). During training, the server uses dialogue data labeled with emotion labels and employs cross-entropy loss for supervised learning.

[0373] 2. Voice Emotion Channel The server extracts acoustic features (such as Mel-frequency cepstral coefficients, fundamental frequency, energy, speech rate, etc.) from digital speech data and inputs them into a convolutional neural network or a temporal neural network for classification. The server uses a multimodal fusion strategy to weightedly fuse the sentiment prediction results of text and speech, thereby obtaining a more accurate understanding of the user's emotional state.

[0374] The server adjusts the content of prompts based on the obtained emotional state. For example, when the emotion is "anger," the server adds instructions such as "Please respond in a more soothing, concise, and clear tone" to the prompt, thereby guiding the generative AI model to produce more suitable text. This adjustment method based on the server's internal technical characteristics improves the human-computer interaction experience and logically establishes a causal chain from feature extraction and model judgment to generation control.

[0375] The server performs risk analysis on the caller information in the security assessment module. The server can use the following non-traditional processing methods: 1. The server maintains a behavioral feature vector for each caller, such as call frequency, call duration distribution, and historical complaint records.

[0376] 2. The server uses clustering or anomaly detection algorithms to compare the feature distribution of new callers with that of historically safe callers. When a new caller deviates significantly from the normal group in the feature space, the server marks it as high-risk.

[0377] 3. The server dynamically generates security assessment prompts based on the risk level, for example: "You are a call security analysis module."

[0378] The current caller's risk level is 'high', and their historical behavioral characteristics are as follows: ... Please specify whether this call should be allowed to proceed, whether functionality needs to be restricted, and any risks that need to be communicated to the user. In this way, the server enables generative AI models to generate security recommendations based on structured features and risk levels. Security logic no longer relies solely on manual rules, but introduces model-driven comprehensive judgment, thereby improving the ability to identify complex risk patterns.

[0379] VI. Communication Interface and Optical Communication Device Control In the communication interface control module, the server abstracts a unified control interface for different types of devices. For regular voice calls, the server maps communication control information to parameters of the communication service provider's API; for optical communication devices, the server maps control information to low-level control parameters such as optical power, level modulation mode, and connection port.

[0380] In one implementation, the server includes fields such as "communication medium type," "required bandwidth," and "latency limit" in the control information output by the generative artificial intelligence model. Based on these fields, the server selects the appropriate physical interface. For example, when high bandwidth and low latency are required, the server prioritizes connecting to the backbone network via optical communication devices; when only low-speed control signaling is needed, the server can send commands through a conventional electrical interface.

[0381] The server maps high-level structured instructions to different physical media through a unified semantic control layer, achieving media independence in technology. Thus, when new communication devices are added, the server only needs to implement a small amount of mapping logic without refactoring the natural language processing part, greatly improving system scalability and reducing the probability of communication load configuration errors.

[0382] VII. Implementation Forms and Technical Effects on the Terminal Side In one embodiment, the terminal uses local speech recognition, while in another embodiment, the raw audio is sent directly to a server or a third-party speech recognition service.

[0383] The local audio processing implemented on the terminal includes echo cancellation, noise suppression, and voice activity detection, thereby reducing the amount of invalid data before it enters the network. Because the terminal locally trims silence segments and compresses audio, the amount of data received by the server is reduced, network communication load decreases, and the waiting time for speech recognition and subsequent processing is shortened.

[0384] After receiving the notification message from the server, the terminal can invoke its local TTS engine to convert the text into speech and play it. The terminal displays the structured results, such as service name, time, location, and confirmation number, through a graphical interface for easy user verification. This dual-channel feedback method technically improves usability and fault tolerance.

[0385] VIII. Specific Usage Examples 1. Example of a user ordering a ride The user says to the terminal, "Please call a taxi to my current location." After the terminal collects audio and performs preprocessing, it sends the audio data to a speech recognition service to obtain text information. The text information and location information are then sent to the server. The server generates a parsed prompt statement, inputs it into a generative artificial intelligence model, and obtains structured instruction information, where the intent is to call a ride, the location is the terminal's location, and the time is immediate. Based on this instruction, the server searches its database, selects a suitable service provider, generates communication control information, and initiates a call through the communication interface. Upon successful call and confirmation, the server generates a result prompt statement, guiding the generative artificial intelligence model to generate a natural language prompt, such as "A vehicle has been called for you, expected to arrive in 10 minutes," and sends it to the terminal.

[0386] 2. Example of a user booking a restaurant The user said, "I've booked an Italian restaurant for two people tonight at 7 pm." The terminal converts speech to text and uploads it. The server generates a prompt based on the text and location information, inputs it into a generative AI model, and parses the intent as a restaurant reservation, time 19:00 on the same day, number of guests 2, and cuisine Italian. The server searches its database for nearby restaurants, filters candidates based on ratings, distance, and availability, and can request ranking suggestions from the generative AI model via another prompt, ultimately determining the target restaurant. The server then generates communication control information, initiates the reservation for that time slot through the communication interface, and generates a natural language prompt to return to the terminal after receiving a confirmation number.

[0387] Through the above implementation, the server, terminal, and user work collaboratively within a unified framework. This invention does not merely automate simple manual operations; rather, it employs generative artificial intelligence models within the computer to drive prompts, structured intermediate representations, multimodal analysis of emotion and risk, and unified communication control mapping. These features improve upon traditional voice appointment systems at three levels: data structure, algorithm flow, and device control, resulting in substantial technical improvements in processing speed, recognition accuracy, resource utilization, security, and scalability.

[0388] use Figure 13 The processing flow is explained.

[0389] Step 1: Users issue voice commands to the terminal.

[0390] Users can express their needs in natural language, such as "call a taxi to my place" or "book an Italian restaurant for two at 7 pm tonight".

[0391] Input: The user's natural speech signal (analog audio signal).

[0392] Output: Raw analog audio signal acquired from the terminal microphone.

[0393] The user's voice is transmitted through the air to the terminal microphone, providing the raw acoustic input for subsequent digitization and recognition.

[0394] Step 2: The terminal converts analog audio signals into digital voice data and performs preprocessing.

[0395] The terminal uses an audio driver to convert the analog audio signal collected by the microphone into digital data (e.g., 16kHz, 16bit PCM) with a certain sampling rate and quantization accuracy, and performs noise reduction, echo cancellation and voice activity detection.

[0396] Input: The analog audio signal from step 1.

[0397] Output: Digital speech data (audio frame sequence) with silence segments and noise removed.

[0398] The terminal calculates the energy and spectral characteristics of each audio frame to determine whether it is a valid speech segment. Non-speech segments are discarded or compressed to reduce the amount of data and improve the accuracy of subsequent speech recognition.

[0399] Step 3: The terminal calls the speech recognition service to convert digital voice data into text information.

[0400] The terminal sends the preprocessed digital voice data to the speech recognition API via the network, or calls the local speech recognition engine to perform acoustic modeling and language decoding on the audio to obtain the corresponding text information.

[0401] Input: The digital voice data output from step 2.

[0402] Output: Text information (text string) corresponding to the user's voice content, such as "Please call a taxi to my current location".

[0403] After receiving the speech recognition result, the terminal parses the returned format (such as JSON), extracts the text fields, and removes redundant confidence scores, retaining only the result with the highest confidence score as input for subsequent processing.

[0404] Step 4: The terminal sends text information and context data to the server.

[0405] The terminal constructs a request message, which includes text information, user identifier, terminal location (such as GPS coordinates), timestamp, and device type, and sends it to the server via a secure transmission protocol.

[0406] Input: The text information output in step 3 and the context data available on the terminal.

[0407] Output: The request message (structured data packet) sent to the server.

[0408] The terminal serializes the data (e.g., converts it into key-value pairs or structured fields) and adds necessary header information at the network layer so that the server can correctly parse the source and content.

[0409] Step 5: The server receives the request and updates the session context.

[0410] The server receives request messages from the terminal through the network interface, deserializes them, extracts text information, user identifiers, and context data, and writes or updates the corresponding user's session context structure.

[0411] Input: The request message output in step 4.

[0412] Output: The updated user session context data structure.

[0413] The server checks if there is an unfinished session for the user. If it does, the new text information is appended to the conversation history list, and the latest location and time are recorded. If it does not exist, a new session entry is created, and the mood and security status fields are initialized.

[0414] Step 6: The server generates parsing-based prompts for generative artificial intelligence models.

[0415] Based on the scenario information in the session context (such as making a reservation, calling a car, or ordering food), the server constructs a prompt statement for the parsing task, embeds the user's original text information into it, and specifies the required output fields and rules.

[0416] Input: The session context (including the latest text information) output from step 5.

[0417] Output: Prompt text for intent recognition and parameter extraction.

[0418] The server explicitly requests the model to extract parameters such as intent, time, location, and number of people in the prompt statement, for example: "You are a natural language parsing module for a booking and ride-hailing system."

[0419] User input: 'Please call a taxi to my office building.' Please extract the intent and parameters, and output the following: intention: Place: time: Number of people: If a parameter is not mentioned, enter 'Not specified'. The server generates the above prompt statement by concatenating strings and filling templates, and ensures that the field names and formats are compatible with the subsequent structured parsing logic.

[0420] Step 7: The server inputs the prompts and text information into the generative artificial intelligence model and obtains the parsing results.

[0421] The server inputs the prompt generated in step 6 along with the user's text information into the generative artificial intelligence model. The model's encoder vectorizes the input and the decoder generates a text result containing intent and parameters.

[0422] Input: The prompt text output in step 6 and the user's text information.

[0423] Output: The parsed text includes information such as intent, time, location, and number of people.

[0424] The server internally reads the model output, for example: Intent: To call a taxi Location: Downstairs from the company building Time: Immediate Number of people: 1 During this process, the server relies on the model's self-attention mechanism to weight key semantic positions, thereby making a selection among multiple candidate semantics and improving the accuracy of parameter extraction.

[0425] Step 8: The server converts the model parsing results into structured instruction information.

[0426] The server performs rule parsing on the output text from step 7, converting text lines such as "intent," "location," "time," and "number of people" into field values ​​and filling them into a predefined data structure. At the same time, the server standardizes the time text and performs geographic parsing on the location text.

[0427] Input: The parsed text output from step 7.

[0428] Output: Structured instruction information object (including fields such as standardized timestamp, geographic location, intent type, number of people, etc.).

[0429] The server calls the time parsing module to convert relative times such as "7 PM tonight" into absolute times; it calls the geocoding service to convert phrases such as "downstairs at the company" and "here" into latitude and longitude coordinates; for missing fields, the server fills in default values ​​or marks them as unspecified according to rules to ensure the integrity of the structured instruction information fields.

[0430] Step 9: The server queries the database based on structured instruction information to obtain the service provider's contact information.

[0431] The server uses a database access module to perform retrieval in the database of storage service provider information based on conditions such as intent, location, and time in the structured instruction information; the server filters and sorts based on fields such as distance, rating, and price.

[0432] Input: The structured instruction information output from step 8.

[0433] Output: A collection of contact information (including contact details and service parameters) for one or more candidate service providers.

[0434] The server executes the query statement and selects records from the data table that match the service area, business hours and service type. The server calculates a comprehensive score for the candidate set according to preset weights, selects the one or more with the highest score, and encapsulates their contact information as an intermediate result.

[0435] Step 10: The server generates selection prompts based on candidate service providers and user preferences (optional).

[0436] When the number of candidate service providers is greater than one, the server constructs a new prompt statement, embedding the candidate list and user preference information into it, and then submits it to the generative artificial intelligence model to provide a recommendation order or selection suggestion.

[0437] Input: The set of candidate service providers output from step 9 and user preferences (such as budget and distance preferences).

[0438] Output: Model input prompts containing the recommended order or reasons, and expected output descriptions.

[0439] The server generates a prompt message similar to the following: "You are a restaurant recommendation assistant."

[0440] User preferences: Italian cuisine, moderate budget, distance no more than 5 kilometers.

[0441] List of candidate restaurants: 1. Restaurant A, rating 4.8, moderate price, 2 km away; 2. Restaurant B, rated 4.5, relatively inexpensive, 1 km away; 3. Restaurant C, rating 4.9, prices are on the high side, distance 4 km.

[0442] Please list the restaurant names in the recommended order and provide a brief reason for your recommendation. This allows the server to make weighted comparisons of candidates based on multidimensional attributes, rather than relying on a single hard-coded rule.

[0443] Step 11: The server invokes a generative artificial intelligence model to determine the final service provider (optional).

[0444] The server inputs the prompt from step 10 into the generative artificial intelligence model to obtain the recommended order or specific selection suggestions, and determines the final service provider record accordingly.

[0445] Input: The prompt message generated in step 10 and the candidate service provider data.

[0446] Output: The identifier and contact information of the selected target service provider.

[0447] The server performs structured parsing on the model output, such as extracting "preferred restaurant name" from the text and then finding the corresponding record in the candidate list. If the object recommended by the model is not in the candidate set, the server ignores the result according to the rules and falls back to the rule-based ranking.

[0448] Step 12: The server generates communication control information and initiates communication through the communication service interface.

[0449] The server constructs communication control information based on structured instruction information and contact information of the target service provider, including parameters such as call type (voice call or data request), target number or address, appointment time, number of people, and location, and then calls the communication service interface.

[0450] Input: The structured instruction information output in step 8 and the service provider contact information output in step 9 or step 11.

[0451] Output: Control commands and call / request initiation results sent to the communication service platform or optical communication device.

[0452] The server maps control information to specific interface parameters, such as setting the called number, adding reservation content (via TTS voice or protocol fields), timeout configuration, etc., and then sends a request to the communication platform; the server receives the status returned by the platform and the order number or reservation confirmation result.

[0453] Step 13: The server generates notification prompts based on the communication results and calls a generative artificial intelligence model to generate user prompts.

[0454] After obtaining the booking or ride-hailing result (including success / failure status, time, number, etc.), the server constructs a prompt statement describing the internal result, which is then used by a generative artificial intelligence model to generate natural language prompt text.

[0455] Input: Communication result data (booking status, confirmation number, time, location, etc.).

[0456] Output: User notification text in natural language format.

[0457] The server generates the following type of prompt: "You are a user-friendly system prompt generation module."

[0458] Internal system results: Service type: Taxi reservation Taxi company: A certain taxi company License plate number: XX1234 Estimated arrival time: 18:10 Please generate a concise and clear Chinese prompt to display to the user. The server inserts the result field into the prompt statement and calls the generative artificial intelligence model to obtain a fluent and natural Chinese sentence, such as "A vehicle has been called for you, license plate number XX1234, expected to arrive at 18:10, please wait at the entrance of the community". Step 14: The server adjusts the notification content based on the user's emotional state (optional).

[0459] The server reads the emotion tags output by the emotion analysis module from the session context, and adds style descriptions to the prompt statements according to the emotion type, so that the generative artificial intelligence model can generate responses that are more suitable for the current emotional state.

[0460] Input: The basic prompts prepared in step 13 and the current emotional state of the user.

[0461] Output: A prompt with emotion-adaptive instructions and the final emotion-adapted notification text.

[0462] For example, the server could add a message like "The user is currently feeling tense; please respond in a calming and concise tone" to the prompt, thus guiding the model to output more gentle text such as "Rest assured, we have arranged a vehicle for you, and it will arrive soon."

[0463] Step 15: The server will send a final notification message to the terminal.

[0464] The server encapsulates the natural language notification text obtained in step 13 or 14 together with the structured results (time, location, confirmation number, etc.) into a response message and sends it to the terminal over the network.

[0465] Input: Natural language notification text and structured result data.

[0466] Output: The response data packet sent to the terminal.

[0467] During this process, the server serializes the data and adds necessary status codes and error message fields so that the terminal can correctly parse and display it after receiving it.

[0468] Step 16: The terminal receives and displays the results, and can also broadcast them via voice.

[0469] The terminal receives the response data packet from the server, parses out the notification text and structured fields, displays the appointment results in the form of a list or card on the display interface, and calls the local TTS engine to convert the notification text into speech and broadcast it to the user.

[0470] Input: The response data packet output in step 15.

[0471] Output: The results displayed on the terminal screen and the voice prompts output by the speaker.

[0472] The terminal displays the service name, time, location, and confirmation number on the interface based on structured fields, and uses TTS to generate an audio stream from the text, which is then output through audio drive, thereby achieving multimodal feedback.

[0473] Step 17: The user decides whether to issue further instructions based on the feedback.

[0474] After viewing the screen or listening to the voice prompt, if users need to modify or cancel their operation, they can issue a new command to the terminal via voice again, such as "Change the time to 7:30" or "Cancel the appointment."

[0475] Input: The result information displayed and broadcast on the terminal.

[0476] Output: New natural language instructions (analog audio signals), proceeding to the next round of processing in step 1.

[0477] Users can iterate and interact multiple times, and the system will repeat the above steps. However, the session context contains historical operation information, which enables the server to automatically associate the previous order when parsing new instructions, thus achieving continuous dialogue and instruction cascading processing.

[0478] Application Example 2 The process flow corresponding to the specific processing in Use Case 2 will be described below. The various parts of the system described below are implemented by the data processing device 12 and the intelligent device 14. In addition, the data processing device 12 is referred to as the "server" and the intelligent device 14 is referred to as the "terminal".

[0479] In existing technologies, communication systems typically only perform simple filtering of incoming calls at the signaling level, failing to combine call content and context for a refined assessment of the risks of unknown calls. This leaves users still vulnerable to malicious communications such as sales pitches or scams. Furthermore, existing voice interaction systems often only use voice recognition results for single business calls, lacking a unified mechanism for generating prompts and linking with generative AI models. This prevents the reuse of intent parsing and dialogue strategies across different business scenarios (such as travel booking, facility reservations, and product ordering), increasing system development complexity and computational resource consumption.

[0480] Furthermore, in existing systems, emotion recognition is often disconnected from business logic. Emotional information is not systematically integrated into service content generation and multi-turn interaction control, making it difficult to automatically adjust responses, tone, and service strategies even when a user is detected to be in a state of tension or anger, resulting in a poor user experience. In optical communication network environments, traditional systems often treat them as transparent links, lacking mechanisms to dynamically optimize the input and output data transmission of generative artificial intelligence models based on real-time conditions such as bandwidth and latency, leading to low resource utilization and high response delays.

[0481] Furthermore, in terms of quality inspection at the manufacturing site, existing technologies for identifying defective products largely rely on fixed image analysis models and threshold judgment logic, lacking collaboration with generative artificial intelligence models. This prevents the automatic generation of structured quality descriptions and natural language reports, hindering operational analysis and human-machine collaborative decision-making. Simultaneously, during long-term operation, the system struggles to promptly utilize user operation records and feedback to adaptively adjust prompt statements and model call parameters, resulting in the generative artificial intelligence model's inference results failing to align with real-world business scenarios and user preferences.

[0482] Therefore, there is an urgent need for an integrated system for communication security control, voice intent parsing, emotion-adaptive services, optical communication resource optimization, and manufacturing quality inspection. By introducing a unified prompt generation and generative artificial intelligence model collaboration mechanism on the server side, the system integrates incoming call risk assessment, structured instruction generation for reservation and ordering tasks, emotion-driven response strategy generation, quality inspection result explanation generation, and network resource adaptive control into the same computing architecture. This reduces computing and communication overhead while improving the overall intelligent decision-making capability and user interaction experience of the system. This is the core issue in the field of computer technology that this invention aims to address.

[0483] The specific processing performed by the specific processing unit 290 of the data processing apparatus 12 in Application Example 2 is achieved by the following means.

[0484] In this invention, the server includes a processing unit for receiving incoming call signals and extracting sender identification information, comparing the sender identification information with stored information to distinguish between known and unknown communications; a processing unit for acquiring unknown communication voice messages and converting them into text information through voice recognition processing, assessing the risks associated with sales pitches or scams based on the text information, and deciding to block, hold, or transfer the call accordingly; a processing unit for acquiring user voice messages and converting request content into text information through voice recognition processing, generating prompt statements based on the request content and status information, and inputting them into a generative artificial intelligence model to obtain structured instruction information related to transportation or facility reservations; a processing unit for controlling an external communication interface to automatically execute outbound call processing for reservation or ordering targets via a communication network based on the structured instruction information; and a processing unit for extracting feature quantities from user voice or images to infer emotional state, generating prompt statements based on the emotional state and request content, and inputting them into a generative artificial intelligence model to obtain information for adjusting response content or service content according to the emotional state. The system includes: a processing unit for handling response indication information; a processing unit for controlling the output of terminal devices based on response indication information to provide feedback to users through display devices or acoustic devices; a processing unit for controlling optical communication devices connected to optical transmission paths, converting optical signals into electrical signals to establish a connection with the communication network, and sending and receiving voice data, text information, and input / output data between the system and the generative artificial intelligence model; a processing unit for acquiring product images from image acquisition devices and preprocessing them, inputting the preprocessed images into the learning model to obtain quality judgment results, and generating control information for controlling the production device to reject defective products based on the quality judgment results; a processing unit for generating prompt statements based on quality judgment results and production status information, inputting them into the generative artificial intelligence model to obtain quality report description information, and recording or outputting it; and a processing unit for acquiring user operation records or evaluation information and updating the prompt statement content or model running parameters for the generative artificial intelligence model accordingly to optimize subsequent response content or service content. This enables the formation of an integrated computing process within the same server architecture, encompassing call risk control, voice intent parsing, emotion-adaptive response, optical communication resource management, and manufacturing quality inspection. Through a unified mechanism for generating prompts and invoking generative artificial intelligence models, it enhances the ability to automatically understand and make decisions in complex communication and business scenarios, reduces network and computing resource consumption, and significantly improves user security and interactive experience, thereby achieving a substantial improvement in the overall performance and intelligence level of computer technology.

[0485] "Incoming call signal" refers to control signals and related information transmitted between communication devices via wired or wireless communication networks to indicate the start, maintenance, or termination of a call request.

[0486] "Sender identification information" refers to number information, address information or other identifying data that can be used to identify the identity of the initiator of the communication, including but not limited to telephone numbers, network addresses or terminal identifiers.

[0487] "Stored information" refers to the record data related to the communication object that is pre-stored in the data storage device, including information such as whitelists, blacklists, historical call records and their risk assessment results.

[0488] "Known communication" refers to communication whose sender identification information matches a record in the stored information, thus being identified as having been registered or evaluated in the system.

[0489] "Unknown communication" refers to communication in which the sender's identification information does not have a matching record in the stored information, or where a record exists but a risk assessment has not been completed and the sender is considered to have an unclear identity.

[0490] "Call voice" refers to the audio signal generated by one or both parties during a communication session, which includes voice data that can be parsed into text content by speech recognition processing.

[0491] "Speech recognition processing" refers to the automatic processing of audio signals into corresponding text information by analyzing and matching signals from spoken conversations or user speech.

[0492] “Textual information” refers to data generated by speech recognition processing or other methods that represents semantic content in the form of character sequences, including natural language text and its structured representation.

[0493] "Risk assessment" refers to the process of classifying and classifying communications as potentially sales pitches, fraudulent activities, or other misconduct based on textual information, sender identification information, and relevant context.

[0494] "Blocking" refers to terminating a communication connection at the communication control layer so that an incoming call cannot continue to establish or maintain a call with the user terminal.

[0495] "Hold" refers to a control method that keeps the communication connection in an intermediate state and does not directly present it to the user before deciding whether to transfer the call to the user terminal.

[0496] "Switching" refers to the operation of switching the communication connection from the server-side control channel to the user terminal after initial processing and risk assessment.

[0497] "User-input voice" refers to natural language voice signals input by users through terminal devices, used to express requests, feedback information, or other instructions.

[0498] "Request content" refers to business demand information that represents the user's intent, which is parsed from the user's spoken voice, including appointments, orders, inquiries, control instructions, etc.

[0499] "Status information" refers to contextual data related to user requests, including environmental information that the system can obtain, such as time, location, user attributes, historical records, and network environment status.

[0500] "Prompt statements" refer to natural language or structured descriptive text generated by the server and input into the generative artificial intelligence model to instruct the model to perform specific analysis, reasoning, or generation tasks.

[0501] "Generative artificial intelligence models" refer to artificial intelligence models trained through machine learning methods that can automatically generate text, structured data, or other output results based on input prompts.

[0502] "Structured instruction information" refers to instruction data organized with predetermined fields or data structures, output by generative artificial intelligence models, which is used to directly drive subsequent communication, control, or business processing.

[0503] "Transportation dispatch" refers to the process of booking or dispatching vehicles, modes of transportation, or other modes of travel for users based on their requests.

[0504] "Facility reservation" refers to the reservation process for using a certain facility, venue, or service resource in advance for a specific time period.

[0505] "External communication interface" refers to the hardware or software interface between the server and external communication networks or external service systems for data interaction, including but not limited to network interface modules, communication protocol stacks, and application interfaces.

[0506] "Outbound call processing" refers to the process of initiating a communication connection from a server or related communication device to an external communication object, including actions such as dialing, session establishment, and message sending.

[0507] "Feature quantity" refers to the numerical representation extracted from user voice, images or other perceptual data to characterize emotional state or identity features.

[0508] "Emotional state" refers to a user's psychological or emotional state at a specific point in time, such as anger, joy, tension, relaxation, etc., which can be inferred through feature quantities and models.

[0509] "Response indication information" refers to structured data output by generative artificial intelligence models based on emotional states and request content, used to guide the system in generating specific response content or service strategies.

[0510] "Terminal equipment" refers to a device installed on the user side for interacting with the server and presenting information to the user or executing control commands, including display devices, audio output devices, and input devices.

[0511] A "display device" refers to an output device used to present visual content such as text, images, and graphical interfaces to users.

[0512] "Acoustic device" refers to an output device used to output voice, prompts or other audio signals to a user, including speakers, headphones, etc.

[0513] "Optical transmission path" refers to the physical transmission channel for data transmission based on optical signals, including fiber optic links and related optical connection components.

[0514] "Optical communication device" refers to a communication device used to transmit and receive optical signals on an optical transmission path and to convert optical signals into electrical signals.

[0515] A "communication network" refers to a network system consisting of multiple communication nodes and their connecting media, used to transmit signaling and service data between different devices.

[0516] "Image acquisition device" refers to equipment used to acquire image data of the object or environment under test, including camera devices, scanning devices, etc.

[0517] "Preprocessing" refers to the steps performed on images or other raw data before they are input into the learning model, such as filtering, denoising, resizing, and enhancement, in order to improve the results of subsequent processing.

[0518] A “learning model” refers to a machine learning or deep learning model that is trained and can process input data by classifying, regressing, or extracting features.

[0519] "Quality judgment result" refers to the judgment data output by the learning model, which indicates whether the product is qualified and the type and severity of its defects.

[0520] "Production equipment" refers to mechanical equipment or automated systems used on the manufacturing site to perform operations such as processing, assembly, handling or sorting.

[0521] "Control information" refers to command data or control signals used to instruct production equipment to perform specific actions (such as rejecting defective products, stopping production, adjusting parameters, etc.).

[0522] "Production status information" refers to data indicating the current operating status of the production line, including production cycle time, equipment status, yield rate, defective product statistics, and other information.

[0523] "Quality report description information" refers to natural language descriptions or structured descriptions generated based on quality judgment results and production status information, intended for human reading or system recording.

[0524] "Operation logs" refer to the operation log data generated by users during the use of the system, including request types, interaction steps, selection results, etc.

[0525] "Evaluation information" refers to explicit or implicit feedback data from users regarding the system's response or service effectiveness, including ratings, text evaluations, and behavioral feedback.

[0526] "Model running parameters" refer to configuration parameters that affect the inference behavior of generative artificial intelligence models, including temperature, output length, threshold settings, weight adjustments, etc.

[0527] In various embodiments of this invention, the server, terminal, and user collaboratively constitute a comprehensive communication and control system based on a generative artificial intelligence model and prompt statements. The server operates in a data center or cloud environment, and the terminal can be a smartphone, a gateway connected to a fixed communication device, an industrial production line control terminal, or a smart terminal with a camera and audio module. The implementation of each functional module is described in detail below, combining hardware and software components.

[0528] I. Overall System Composition A server comprises a computing device, a storage device, and a network interface. The computing device may employ a multi-core central processing unit and an optional graphics processing unit. The storage device includes main memory and persistent storage, used to store programs, model parameters, and business data. The network interface is used to connect to the Internet or a dedicated network and to communicate with terminals and external service systems.

[0529] The server runs an operating system (such as a Unix-like operating system) at the software level, deploys an application server framework (such as an HTTP / HTTPS-based service framework), and installs the following software components: a speech recognition service client (such as a general cloud speech recognition API client), a natural language processing library (such as a general word segmentation and entity recognition library), a machine learning framework (such as TensorFlow, a general deep learning framework), a database management system (such as a relational database), communication control software (such as a session control module, a softswitch module), and a generative artificial intelligence model access interface.

[0530] The terminal includes a processor, memory, an audio acquisition module (microphone), an audio output module (speaker or headphones), an optional image acquisition module (camera), a display module, and a network communication module. The terminal runs a lightweight client program and interacts with the server via application layer protocols.

[0531] Users input data via voice, video, or user interface through the terminal. The server executes a series of specific data structures and algorithms on the data received over the network to achieve functions such as call security assessment, intent parsing, emotion-adaptive services, and manufacturing quality control.

[0532] II. Initial Call Handling and Risk Assessment Forms In fixed-line communication scenarios, the server connects to fixed-line communication devices or VoIP gateways via communication control software. The server is configured with call control logic so that all incoming calls first enter the initial processing module on the server side.

[0533] After receiving an incoming call, the server extracts the sender identification information from the signaling header fields and standardizes it into a uniform format (such as a numeric string or a Uniform Resource Identifier). The server uses this identification information as a key to perform an index lookup in the "Communication Object Table" of the database. This table records fields such as sender identification, category label (whitelist, blacklist, or unknown), historical call count, and cumulative risk score. Based on the query results, the server determines whether the incoming call belongs to known or unknown communication.

[0534] When the server detects unknown communication, it controls the communication channel to enter automatic answer mode and collects the call audio from the voice channel. The server buffers the audio signal in segments according to a fixed frame length and sampling rate, and calls the voice recognition service client to compress and encode the audio signal before uploading it to the cloud-based voice recognition service. The server receives the recognition result text and stores it in the "Call Content Table" in a unified text field format.

[0535] The server performs natural language processing on the text, extracting keywords, tone features, and semantic patterns. For example, the server assigns higher risk weights to text containing high-risk words such as "limited time," "investment," and "verification code" at the rule level. Simultaneously, the server uses the processed text and rule output as input features to construct a prompt statement, which is then sent to the generative AI model. This prompt statement could be: "Below is the content of a call from an unknown number. Please determine whether the call is a normal business contact, a sales / advertisement, or a suspected scam, and explain your reasoning. The call content is as follows: '..." The server obtains category labels and risk levels based on the output of the generative artificial intelligence model. It then weights and fuses the categories and risk levels with the results from the rule engine to arrive at a final risk assessment. Based on this assessment, the server controls communication devices to perform blocking, hold, or transfer operations. Because the server utilizes the model to perform multi-dimensional abstract representations of natural language (e.g., contextual attention weights) and pattern recognition, it can structure more implicit semantic patterns at the computational layer compared to relying solely on fixed keyword matching, thereby improving risk identification accuracy and reducing the false positive rate.

[0536] III. Voice Request Parsing and Reservation / Order Control Mode After receiving voice messages from users on the terminal, the server performs audio preprocessing (including noise reduction and silence detection) and then converts them into text information through a speech recognition service. The server uses a natural language processing library to segment the text, tag parts of speech, and recognize entities, extracting the request content into an intermediate structure, such as fields for request type (travel, food ordering, etc.), time information, and location information.

[0537] The server generates a prompt statement based on this, writing the user's original message, extraction results, and relevant status information (such as user location, historical preferences, etc.) into the prompt statement. An example prompt statement could be: "The user says: 'Please book me a taxi to the airport at 7 PM tonight.' Given: The user's current location is No. 100, a street in a certain city. Please extract the user's intent and output the service type, departure time, departure location, and destination." The server sends the prompts to the generative AI model interface, which typically employs an encoder-decoder structure with attention mechanisms. The input side converts the prompts into word vectors, which are then processed through multiple layers of self-attention and feedforward networks to form contextual feature representations. The output side maps the results to structured text, such as natural language segments containing field names and values, according to task requirements.

[0538] The server parses structured instruction information from the text returned by the model. For example, the server parses the content such as "Service type: taxi; Departure time: 19:00 today; Departure location: No. 100, a certain street; Destination: Terminal 3 of a certain airport" in the generated result and converts it into an internal data structure to drive the external communication interface.

[0539] The server controls the communication interface module to send network requests to external travel service systems, or automatically dials for reservations by calling the telephone control interface through session control software. Compared to the traditional method of developers manually writing rules to map natural language to fixed API parameters, this invention dynamically generates structured instruction information through a generative artificial intelligence model. This enables the server to handle more complex and varied natural language expressions, while reducing the need to write hard-coded rules for each expression. As a result, the complexity of conditional branches in the program structure is reduced, and scalability is improved.

[0540] IV. Emotion Recognition and Emotion Adaptive Service Models When processing user voice or image data, the server extracts various features from the signal as input to the emotion recognition model. Voice features may include pitch trajectory, energy envelope, speech rate, formant distribution, etc.; image features may include facial key point positions, muscle texture changes, etc. The server inputs these feature vectors into a pre-trained emotion classification model, which can employ convolutional and recurrent structures or multi-head self-attention structures to output probability distributions for different emotion labels.

[0541] The server combines the identified emotion tag with the user's request content to create a prompt message. For example, when a user's voice message is "Why is this bill so expensive?" and the emotion tag is anger, the server generates the following prompt message: "The user says, 'Why is this bill so expensive?' Sentiment analysis indicates that the user is in a state of anger. Please generate a polite, patient, and soothing response, no more than 80 characters." The server sends the prompt to a generative artificial intelligence model. During pre-training, the model has learned language patterns for different tones, levels of politeness, and contexts. The server uses the model's output to generate a customer service response tailored to the current emotional state. The server then presents this response text to the user via the terminal.

[0542] The server records user feedback information during multiple rounds of interaction, such as whether the user continues to express negative emotions. Based on the feedback data, the server adjusts the structure and content of the prompts over time, for example, adding safety reminders for high-risk situations and buffering explanations for high-pressure emotions. In this way, the present invention ensures that emotion recognition information is not only stored at the label output layer, but is systematically embedded in the design of prompts and language generation strategies, effectively improving the quality of interaction.

[0543] V. Optimization of Data Transmission Forms in Optical Communication and Generative Artificial Intelligence When a server connects to an optical communication device, it obtains operating parameters such as link bandwidth, current load, and round-trip time through optical physical layer and link layer signaling. The server updates this network status information periodically or when the load changes and stores it in a network status table.

[0544] When invoking a generative AI model, the server adjusts data block size, compression strategy, and transmission frequency based on the current network conditions. For example, when increased latency or reduced bandwidth is detected, the server chooses an encoding method with a higher compression ratio, reduces redundant information in prompts, or uses segmented interaction to reduce the size of a single request. The server can generate prompts describing network constraints, explicitly indicating output length limitations, thereby enabling the generative AI model to produce shorter response text and reducing network load.

[0545] Because the server incorporates network status as explicit fields into the prompt statements and drives the length and content style of the model output, this invention implements a network load control method based on model behavior adjustment at the computing and communication layers. Compared with traditional schemes that rely solely on transport layer congestion control, this method can more directly reduce unnecessary data volume, thereby reducing latency and improving the overall system response speed.

[0546] VI. Image Quality Detection and Description of Generated Forms In a manufacturing setting, the server acquires product images using industrial cameras. The server preprocesses the raw images using an image processing library, performing actions such as grayscale conversion, noise reduction, contrast enhancement, and geometric normalization to convert the images into a unified tensor format. The server then inputs the preprocessed images into a quality detection model based on a convolutional neural network. During the training phase, this model uses a set of images of both good and defective products for supervised learning, optimizing the convolutional kernel weights through backpropagation.

[0547] The server obtains the probability of various defects from the model output. Based on preset thresholds, the server classifies products as good or defective, and further identifies defect types, such as scratches or stains. According to the judgment results, the server sends control commands to the production unit, driving the actuators via the industrial control bus to remove defective products from the production line.

[0548] When generating a quality report, the server combines the model's output with production status information to construct a prompt statement. For example: Product type: Phone case. Inspection result: Defective. Defect type: Surface scratches, area: upper right corner. Please describe this quality problem and the corrective measures taken in Chinese, in no more than 100 characters. The server sends the prompt to a generative artificial intelligence model, which generates a natural language explanation. The server then stores this explanation text in a quality log table and displays it on the management terminal interface. By transforming deep convolutional features and statistical information into readable instructions, the server conveniently provides explanatory information to human operators, helping to quickly locate problematic processes and improve production parameters.

[0549] VII. Joint Optimization Form of Prompt Statements and Generative Artificial Intelligence Models During long-term operation, the server records the input prompts, output results, and deviations from the actual execution results for each call to the generative artificial intelligence model. The server uses these logs to build training samples, correlating the prompts, predictions, and expected results to fine-tune model parameters or optimize prompt templates.

[0550] For example, the server can analyze which prompt statement structures are likely to cause the model to misunderstand the user's intent, and then automatically adjust the order of the prompt statements, add explanations of key fields, or constrain the output format. The server can also generate multiple prompt statement templates for different business scenarios and select the one that performs best in the current scenario through a statistical model, thereby achieving adaptive selection of prompt statements.

[0551] This structured management and feedback loop for prompt statements transforms the use of generative AI models from a simple "black box invocation" into an optimizable and adjustable prompt design system at the server layer. Through joint learning of prompt statements and model parameters, the server gradually reduces the deviation between model output and business expectations, improving the overall system's processing accuracy.

[0552] VIII. Multiple Implementation Forms and Alternative Structures The server can use different specific software components in different implementations. For example, speech recognition can use cloud services from different vendors or locally trained acoustic models, while generative artificial intelligence models can use pre-trained language models or domain-specific models with different architectures. The database system can use relational or non-relational structures, but the server maintains a unified intermediate data structure design and prompt generation mechanism in all implementations.

[0553] The terminal can be a smartphone, tablet, IP adapter connected to a landline phone, industrial control panel, or smart display with a camera and speaker in different scenarios. As long as the terminal can perform audio and video acquisition and network communication, it can work in collaboration with the server through the program of this invention.

[0554] Users can be individual users, call center agents, manufacturing site operators, or system maintenance personnel in different application scenarios. The server adds different contextual information and constraints to the prompts based on the user's role and permissions, thus ensuring that the generative AI model outputs information that meets the needs of the corresponding role.

[0555] Through the above-mentioned multiple implementation forms, this invention introduces a unified prompt statement generation and generative artificial intelligence model collaborative framework within the server, integrating incoming call risk assessment, voice intent parsing, emotion adaptation, network resource optimization, and quality detection into a closely coupled computing process. This improves the accuracy and efficiency of information processing from multiple aspects, including data structure, model mechanism, and communication control, thereby achieving a substantial improvement in computer technology itself.

[0556] use Figure 14 The processing flow is explained.

[0557] Step 1: Users send voice requests through their terminals.

[0558] Users can speak natural language to the terminal, such as "Help me book a taxi to the airport at 7 pm tonight".

[0559] Input: The user's natural language speech.

[0560] Output: The raw audio data stream buffered internally by the terminal.

[0561] The terminal acquires continuous audio signals through a microphone, converts the analog signals into digital PCM data through a sound card, and writes them into a memory buffer at fixed frame lengths (e.g., 20ms per frame) to prepare for subsequent encoding and transmission.

[0562] Step 2: The terminal encodes the audio and sends it to the server.

[0563] The terminal reads PCM audio data from memory, uses a built-in audio encoder (such as an encoding module based on a general compression algorithm) to compress the high bit rate PCM data into lower bit rate audio frames, and assembles these compressed audio frames into HTTP or WebSocket packets.

[0564] Input: PCM audio data cached in the terminal.

[0565] Output: Compressed audio data packets sent over the network.

[0566] The terminal constructs a request message at the application layer, adds metadata such as user identifier, device identifier, and timestamp to the message header, and transmits it through the network module (Wi-Fi). (Fi / Ethernet / Cellular) sends data packets to the server's designated interface.

[0567] Step 3: The server receives audio data and performs preprocessing for speech recognition.

[0568] After receiving data packets from the terminal at the network interface, the server parses the compressed audio payload from the application layer and reconstructs the audio frame sequence in memory. The server performs simple preprocessing on the audio, such as energy detection, silence removal, and normalization, to improve subsequent recognition results.

[0569] Input: Compressed audio data packets from the terminal.

[0570] Output: Preprocessed consecutive audio frames.

[0571] The server decodes the compressed audio into PCM format using a decoding module, splices multiple audio frames in chronological order, and sends the speech segments without silence to the speech recognition API client.

[0572] Step 4: The server calls the speech recognition service to generate text information.

[0573] The server uses a speech recognition service client to package preprocessed audio frames and configuration parameters (sampling rate, language code, etc.) into a recognition request, which is then sent to the cloud-based speech recognition service via HTTPS. The cloud service performs feature extraction and acoustic / language model inference on the audio, returning the corresponding text.

[0574] Input: Preprocessed audio frames and recognition configuration parameters.

[0575] Output: Text information representing the user's message.

[0576] The server parses and identifies the response message, extracts the optimal transcribed string, such as "Book me a taxi to the airport at 7 pm tonight", and stores it in the session context data structure for subsequent natural language processing and generative artificial intelligence model calls.

[0577] Step 5: The server performs natural language parsing on the text information and constructs an intermediate semantic structure.

[0578] The server uses a natural language processing library to perform word segmentation, part-of-speech tagging, time expression parsing, and location entity recognition on text information, converting sentences into an intermediate representation that includes fields such as intent category, time slot, and location slot.

[0579] Input: Text information obtained from user speech transcription.

[0580] Output: An intermediate semantic structure containing intent and slot information.

[0581] The server identifies key actions (e.g., "book"), target objects (e.g., "taxi"), time ("7 PM tonight"), implicit departure location (defaulting to the user's current location), and destination ("airport") based on parts of speech and dependency relationships, and encodes this information into structured data, such as intent type fields and multiple parameter fields.

[0582] Step 6: The server generates prompts based on the intermediate structure and invokes the generative artificial intelligence model.

[0583] The server integrates the user's original sentence, intermediate semantic structure, and relevant state information (such as the user's geographical location and historical preferences) into a prompt statement, describing the task that needs to be performed by the generative artificial intelligence model in natural language.

[0584] Input: intermediate semantic structure, user state information.

[0585] Output: Prompt text used to drive generative artificial intelligence models.

[0586] The server constructs an example of a prompt statement: "The user says: 'Book me a taxi to the airport at 7 PM tonight.' Known information: The user's current location is No. 100, a certain street in a certain city. Please extract the user's intent and output the service type, departure time, departure location, and destination." The server passes this prompt statement to the generative artificial intelligence model interface, and attaches the model type and generation parameters (such as temperature and maximum character count) to the request.

[0587] Step 7: The server obtains structured instruction information from the generative artificial intelligence model.

[0588] After receiving prompts, the generative AI model internally maps each word to a vector through an embedding layer. This vector is then transformed into a contextual representation through multiple layers of self-attention and feedforward networks. Finally, at the output, it generates text containing structured elements according to the instructions. The server receives the model's output.

[0589] Input: A text message containing task descriptions and context.

[0590] Output: A text result containing structured instructions.

[0591] The server parses the model's output text, extracts field values ​​(e.g., "Service type: taxi; Departure time: 19:00 today; Departure location: No. 100, a certain road; Destination: Terminal 3 of a certain airport"), and loads the results into the order task object inside the system as the basis for subsequent external communication and control.

[0592] Step 8: The server uses structured instruction information to control the external communication interface to complete the reservation or order placement.

[0593] The server reads fields such as service type, time, and location from the order task object and queries the service provider information table for matching service endpoints (such as the nearest travel service platform or catering service interface).

[0594] Input: Structured instruction information and server endpoint database.

[0595] Output: Reservation / order requests sent to external service systems and the corresponding response results.

[0596] The server constructs an HTTP request or initiates a telephone session according to the protocol requirements of the server endpoint, sends the request through the external communication interface module, and receives the response from the other party, such as the order number and estimated arrival time. The server writes this response data into the internal order status table and updates the task status.

[0597] Step 9: The server performs emotion recognition and generates emotion-sensitive response strategies.

[0598] The server extracts feature vectors from the user's original speech or images, inputs them into an emotion recognition model, and obtains emotion labels and confidence scores. Subsequently, based on the emotion labels and the user's request content, the server generates new prompts, requiring the generative AI model to create a response or service strategy suitable for the current emotional state.

[0599] Input: User's voice / image features and identified emotion tags.

[0600] Output: Emotionally sensitive response text or service policy instruction information.

[0601] The server can construct a prompt message such as: "The user says: 'Why is this bill so expensive?' Sentiment analysis indicates the user is in a state of anger. Please generate a polite, patient, and soothing response, no more than 80 characters." The generative AI model outputs the corresponding response, which the server returns to the terminal as part of the response content.

[0602] Step 10: The terminal displays the results and prompts returned by the server to the user.

[0603] The terminal receives order confirmation information, emotionally sensitive response text, and other status data from the server, and presents this information to the user through the display module or audio output module.

[0604] Input: Order details, reply text, and status code returned by the server.

[0605] Output: A user-perceptible visual interface or voice broadcast.

[0606] The terminal displays text on the screen such as "A taxi has been booked for you to arrive at the airport at 19:00 tonight, license plate number ×××, expected arrival at 18:45," and calls the local TTS engine to convert the text into speech, which is then played to the user through the speaker, thus completing a full request processing loop.

[0607] The specific processing unit 290 sends 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 sound representing user input regarding the result of the specific processing. The control unit 46A sends the sound data representing 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 sound data.

[0608] Data generation model 58 is a so-called generative AI (Artificial Intelligence). Examples of data generation models 58 include ChatGPT (registered trademark) (accessible via the internet (URL: https: / / openai.com / blog / chatgpt)). Data generation model 58 is obtained through deep learning on a neural network. Input to data generation model 58 are prompt words containing instructions, and inference data such as sound data representing sound, text data representing text, and image data representing images (e.g., still image data or animation data). Data generation model 58 infers from the input inference data based on the instructions represented by the prompt words and outputs the inference result in one or more data forms, such as sound data, text data, and image data. Data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or induction. The specific processing unit 290 performs the aforementioned specific processing while using data generation model 58. The data generation model 58 can also be a model finely tuned to output inference results from prompts that do not contain instructions. In this case, the data generation model 58 can output inference results based on prompts that do not contain instructions. The data processing apparatus 12, etc., includes various data generation models 58, including AI other than the generation AI. AI other than the generation AI can be, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and can perform various processes, but is not limited to this example. Furthermore, the AI ​​can also be an AI agent. Furthermore, when the processing of the above-mentioned parts is performed by AI, the processing can be performed partially or entirely by AI, but is not limited to this example. Furthermore, the processing performed by the AI ​​including the generation AI can be replaced by processing in the rule base, and the processing in the rule base can also be replaced by processing performed by the AI ​​including the generation AI.

[0609] Furthermore, the processing of the aforementioned data processing system 10 is performed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart device 14, but it can also be performed by both the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart device 14. Additionally, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the smart device 14 or external devices, and the smart device 14 acquires or collects information required for processing from the data processing device 12 or external devices.

[0610] For example, the collection unit is implemented by the control unit 46A of the smart device 14 or the specific processing unit 290 of the data processing device 12. For example, the acquisition unit uses the camera 42 or communication I / F 44 of the smart device 14 to acquire step data, which is then processed by the specific processing unit 290 of the data processing device 12. For example, the analysis unit is implemented by the specific processing unit 290 of the data processing device 12, which analyzes the data from the collection unit and the acquisition unit. For example, the generation unit is implemented by the specific processing unit 290 of the data processing device 12, which uses a generation AI to generate a menu. For example, the serving unit is implemented by the output device 40 of the smart device 14 or the specific processing unit 290 of the data processing device 12, which provides the generated menu to the user. The correspondence between each unit and the device or control unit is not limited to the above examples and various changes can be made.

[0611] In the above embodiments, examples of specific processing by the data processing device 12 are given, but the technology disclosed herein is not limited to this, and specific processing may also be performed by the smart device 14.

[0612] Second Implementation Method Figure 3 An example of the configuration of the data processing system 210 according to the second embodiment is shown.

[0613] like Figure 3 As shown, the data processing system 210 includes a data processing device 12 and smart glasses 214. A server can be cited as an example of the data processing device 12.

[0614] The data processing apparatus 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" as understood in this disclosure. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0615] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, camera 42, and communication I / F 44 are also connected to the bus 52.

[0616] Microphone 238 receives instructions from user 20 by receiving sounds emitted by user 20. Microphone 238 captures sounds emitted by user 20 and converts the captured sounds into sound data, which is then output to processor 46. Speaker 240 outputs sound according to instructions from processor 46.

[0617] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and imaging elements such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, which captures images of the user 20's surroundings (e.g., the field of view defined by an angle equivalent to the field of vision of an average healthy person).

[0618] Communication I / F44 is connected to network 54. Communication I / F44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F44 and 26 is performed in a secure state.

[0619] Figure 4 This illustrates an example of the main functions of the data processing device 12 and the smart glasses 214. For example... Figure 4 As shown, in the data processing device 12, specific processing is performed by the processor 28. The specific processing program 56 is stored in the memory 32.

[0620] The specific processing program 56 is an example of a "program" involved in the technology of this disclosure. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0621] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. In the emotion inference function (emotion-specific function) using the emotion-specific model 59, various inferences and predictions related to the user's emotions are performed, including inferences and predictions of the user's emotions, but this is not limited to this example. Furthermore, emotion inference and prediction may also include, for example, emotion analysis (parsing).

[0622] In the smart glasses 214, the processor 46 performs reception and output processing. The memory 50 stores the reception and output program 60. The processor 46 reads the reception and output program 60 from the memory 50 and executes the read reception and output program 60 on the RAM 48. The reception and output processing is implemented by the processor 46 operating as a control unit 46A according to the reception and output program 60 executed on the RAM 48. Furthermore, the smart glasses 214 has the same data generation model and emotion-specific model as the data generation model 58 and the emotion-specific model 59, and these models can also be used to perform the same processing as the specific processing unit 290.

[0623] Next, the specific processing of the specific processing unit 290 of the data processing device 12 will be described. Each part of the system described below is implemented by the data processing device 12 and the smart glasses 214. 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".

[0624] Example 1 The process is the same as that of the specific process described in Embodiment 1 in the first embodiment above, so the description is omitted.

[0625] Application Example 1 The process is the same as that in the specific processing described in Application Example 1 of the first embodiment above, so the description is omitted.

[0626] Example 2 The process is the same as that of the specific process in Embodiment 2 described in the first embodiment above, so the description is omitted.

[0627] Application Example 2 The process is the same as that in the specific processing described in Application Example 2 of the first embodiment above, so the description is omitted.

[0628] The specific processing unit 290 sends the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A outputs the result of the specific processing to the speaker 240. The microphone 238 acquires sound input representing the user's input regarding the result of the specific processing. The control unit 46A sends the sound data representing the user's input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the sound data.

[0629] Data generation model 58 is a so-called generative AI (Artificial Intelligence). Examples of data generation models 58 include ChatGPT (registered trademark) (accessible via the internet (URL: https: / / openai.com / blog / chatgpt)). Data generation model 58 is obtained through deep learning on a neural network. Input to data generation model 58 are prompt words containing instructions, and inference data such as sound data representing sound, text data representing text, and image data representing images (e.g., still image data or animation data). Data generation model 58 infers from the input inference data based on the instructions represented by the prompt words and outputs the inference result in one or more data forms, such as sound data, text data, and image data. Data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or induction. The specific processing unit 290 performs the aforementioned specific processing while using data generation model 58. The data generation model 58 can also be a model finely tuned to output inference results from prompts that do not contain instructions. In this case, the data generation model 58 can output inference results based on prompts that do not contain instructions. The data processing apparatus 12, etc., includes various data generation models 58, including AI other than the generation AI. AI other than the generation AI can be, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and can perform various processes, but is not limited to this example. Furthermore, the AI ​​can also be an AI agent. Furthermore, when the processing of the above-mentioned parts is performed by AI, the processing can be performed partially or entirely by AI, but is not limited to this example. Furthermore, the processing performed by the AI ​​including the generation AI can be replaced by processing in the rule base, and the processing in the rule base can also be replaced by processing performed by the AI ​​including the generation AI.

[0630] Furthermore, the processing of the aforementioned data processing system 10 is performed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart glasses 214, but it can also be performed by both the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart glasses 214. Additionally, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the smart glasses 214 or external devices, and the smart glasses 214 acquires or collects information required for processing from the data processing device 12 or external devices.

[0631] For example, the collection unit is implemented by the control unit 46A of the smart glasses 214 or the specific processing unit 290 of the data processing device 12. For example, the acquisition unit uses the camera 42 or communication I / F 44 of the smart glasses 214 to acquire step data, which is then processed by the specific processing unit 290 of the data processing device 12. For example, the analysis unit is implemented by the specific processing unit 290 of the data processing device 12, which analyzes the data from the collection unit and the acquisition unit. For example, the generation unit is implemented by the specific processing unit 290 of the data processing device 12, which uses a generation AI to generate a menu. For example, the serving unit is implemented by the speaker 240 of the smart glasses 214 or the specific processing unit 290 of the data processing device 12, which provides the generated menu to the user. The correspondence between each unit and the device or control unit is not limited to the above examples and various changes can be made.

[0632] In the above embodiments, examples of specific processing by the data processing device 12 are given, but the technology disclosed herein is not limited to this, and specific processing may also be performed by the smart glasses 214.

[0633] Third Implementation Method Figure 5 An example of the configuration of the data processing system 310 according to the third embodiment is shown.

[0634] like Figure 5 As shown, the data processing system 310 includes a data processing device 12 and a head-mounted terminal 314. A server can be cited as an example of the data processing device 12.

[0635] The data processing apparatus 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" as understood in this disclosure. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0636] The head-mounted terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, camera 42, display 343, and communication I / F 44 are also connected to the bus 52.

[0637] Microphone 238 receives instructions from user 20 by receiving sounds emitted by user 20. Microphone 238 captures sounds emitted by user 20 and converts the captured sounds into sound data, which is then output to processor 46. Speaker 240 outputs sound according to instructions from processor 46.

[0638] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and imaging elements such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, which captures images of the user 20's surroundings (e.g., the field of view defined by an angle equivalent to the field of vision of an average healthy person).

[0639] Communication I / F44 is connected to network 54. Communication I / F44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F44 and 26 is performed in a secure state.

[0640] Figure 6 This illustrates an example of the main functions of the data processing device 12 and the head-mounted terminal 314. For example... Figure 6 As shown, in the data processing device 12, specific processing is performed by the processor 28. The specific processing program 56 is stored in the memory 32.

[0641] The specific processing program 56 is an example of a "program" involved in the technology of this disclosure. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0642] The memory 32 stores the data generation model 58 and the emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290.

[0643] In the head-mounted terminal 314, the processor 46 performs the acceptance / output processing. The memory 50 stores the acceptance / output program 60. The processor 46 reads the acceptance / output program 60 from the memory 50 and executes the read acceptance / output program 60 on the RAM 48. The acceptance / output processing is implemented by the processor 46 operating as a control unit 46A according to the acceptance / output program 60 executed on the RAM 48.

[0644] Next, the specific processing of the specific processing unit 290 of the data processing device 12 will be described. Each part of the system described below is implemented by the data processing device 12 and the head-mounted terminal 314. In the following description, the data processing device 12 will be referred to as the "server" and the head-mounted terminal 314 will be referred to as the "terminal".

[0645] Example 1 The process is the same as that of the specific process described in Embodiment 1 in the first embodiment above, so the description is omitted.

[0646] Application Example 1 The process is the same as that in the specific processing described in Application Example 1 of the first embodiment above, so the description is omitted.

[0647] Example 2 The process is the same as that of the specific process in Embodiment 2 described in the first embodiment above, so the description is omitted.

[0648] Application Example 2 The process is the same as that in the specific processing described in Application Example 2 of the first embodiment above, so the description is omitted.

[0649] The specific processing unit 290 sends the result of the specific processing to the head-mounted terminal 314. In the head-mounted terminal 314, the control unit 46A outputs the result of the specific processing to the speaker 240 and the display 343. The microphone 238 acquires sound input representing the user's input regarding the result of the specific processing. The control unit 46A sends the sound data representing the user's input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the sound data.

[0650] Data generation model 58 is a so-called generative AI (Artificial Intelligence). Examples of data generation models 58 include ChatGPT (registered trademark) (accessible via the internet (URL: https: / / openai.com / blog / chatgpt)). Data generation model 58 is obtained through deep learning on a neural network. Input to data generation model 58 includes prompt words containing instructions, and inference data such as sound data representing sound, text data representing text, and image data representing images (e.g., still image data or animation data). Data generation model 58 infers the input inference data based on the instructions represented by the prompt words and outputs the inference result in one or more data forms such as sound data, text data, and image data. Data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or induction. The specific processing unit 290 performs the aforementioned specific processing while using data generation model 58. The data generation model 58 can also be a model finely tuned to output inference results from prompts that do not contain instructions. In this case, the data generation model 58 can output inference results based on prompts that do not contain instructions. The data processing apparatus 12, etc., includes various data generation models 58, including AI other than the generation AI. AI other than the generation AI can be, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and can perform various processes, but is not limited to this example. Furthermore, the AI ​​can also be an AI agent. Furthermore, when the processing of the above-mentioned parts is performed by AI, the processing can be performed partially or entirely by AI, but is not limited to this example. Furthermore, the processing performed by the AI ​​including the generation AI can be replaced by processing in the rule base, and the processing in the rule base can also be replaced by processing performed by the AI ​​including the generation AI.

[0651] Furthermore, the processing of the aforementioned data processing system 10 is performed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the head-mounted terminal 314, but it can also be performed by both the specific processing unit 290 of the data processing device 12 and the control unit 46A of the head-mounted terminal 314. Additionally, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the head-mounted terminal 314 or external devices, and the head-mounted terminal 314 acquires or collects information required for processing from the data processing device 12 or external devices.

[0652] For example, the collection unit is implemented by the control unit 46A of the head-mounted terminal 314 or the specific processing unit 290 of the data processing device 12. For example, the acquisition unit uses the camera 42 or communication I / F 44 of the head-mounted terminal 314 to acquire step data, which is then processed by the specific processing unit 290 of the data processing device 12. For example, the analysis unit is implemented by the specific processing unit 290 of the data processing device 12, which analyzes the data from the collection unit and the acquisition unit. For example, the generation unit is implemented by the specific processing unit 290 of the data processing device 12, which uses a generation AI to generate a menu. For example, the serving unit is implemented by the speaker 240 and display 343 of the head-mounted terminal 314 or the specific processing unit 290 of the data processing device 12, which provides the generated menu to the user. The correspondence between each unit and the device or control unit is not limited to the above examples and various changes can be made.

[0653] In the above embodiments, examples of specific processing by the data processing device 12 are given, but the technology disclosed herein is not limited to this, and specific processing may also be performed by the head-mounted terminal 314.

[0654] Fourth Implementation Method Figure 7 An example of the configuration of the data processing system 410 according to the fourth embodiment is shown.

[0655] like Figure 7 As shown, the data processing system 410 includes a data processing device 12 and a robot 414. A server can be cited as an example of the data processing device 12.

[0656] The data processing apparatus 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" as understood in this disclosure. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0657] Robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a controlled object 443. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, camera 42, controlled object 443, and communication I / F 44 are also connected to the bus 52.

[0658] Microphone 238 receives instructions from user 20 by receiving sounds emitted by user 20. Microphone 238 captures sounds emitted by user 20 and converts the captured sounds into sound data, which is then output to processor 46. Speaker 240 outputs sound according to instructions from processor 46.

[0659] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and imaging elements such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, to photograph the area around robot 414 (e.g., the field of view defined by a perspective equivalent to the field of vision of an average healthy person).

[0660] Communication I / F44 is connected to network 54. Communication I / F44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F44 and 26 is performed in a secure state.

[0661] The controlled object 443 includes a display device, LEDs (light-emitting diodes) for the eyes, and motors for driving the arms, hands, and feet. The posture or movement of the robot 414 is controlled by controlling the motors in the arms, hands, and feet. Some of the robot 414's emotions can be expressed by controlling these motors. In addition, the robot 414's facial expressions can also be expressed by controlling the illumination state of the LEDs in its eyes.

[0662] Figure 8 This illustrates an example of the main functions of the data processing device 12 and the robot 414. For example... Figure 8 As shown, in the data processing device 12, specific processing is performed by the processor 28. The specific processing program 56 is stored in the memory 32.

[0663] The specific processing program 56 is an example of a "program" involved in the technology of this disclosure. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0664] The memory 32 stores the data generation model 58 and the emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290.

[0665] In robot 414, the processor 46 performs the acceptance and output processing. The memory 50 stores the acceptance and output program 60. The processor 46 reads the acceptance and output program 60 from the memory 50 and executes the read acceptance and output program 60 on RAM 48. The acceptance and output processing is implemented by the processor 46 acting as the control unit 46A according to the acceptance and output program 60 executed on RAM 48.

[0666] Next, the specific processing of the specific processing unit 290 of the data processing device 12 will be described. Each part of the system described below is implemented by the data processing device 12 and the robot 414. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 will be referred to as the "terminal".

[0667] Example 1 The process is the same as that of the specific process described in Embodiment 1 in the first embodiment above, so the description is omitted.

[0668] Application Example 1 The process is the same as that in the specific processing described in Application Example 1 of the first embodiment above, so the description is omitted.

[0669] Example 2 The process is the same as that of the specific process in Embodiment 2 described in the first embodiment above, so the description is omitted.

[0670] Application Example 2 The process is the same as that in the specific processing described in Application Example 2 of the first embodiment above, so the description is omitted.

[0671] The specific processing unit 290 sends the result of the specific processing to the robot 414. In the robot 414, the control unit 46A outputs the result of the specific processing to the speaker 240 and the controlled object 443. The microphone 238 acquires sound input representing the result of the specific processing. The control unit 46A sends the sound data representing the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the sound data.

[0672] Data generation model 58 is a so-called generative AI (Artificial Intelligence). Examples of data generation models 58 include ChatGPT (registered trademark) (accessible via the internet (URL: https: / / openai.com / blog / chatgpt)). Data generation model 58 is obtained through deep learning on a neural network. Input to data generation model 58 are prompt words containing instructions, and inference data such as sound data representing sound, text data representing text, and image data representing images (e.g., still image data or animation data). Data generation model 58 infers from the input inference data based on the instructions represented by the prompt words and outputs the inference result in one or more data forms, such as sound data, text data, and image data. Data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or induction. The specific processing unit 290 performs the aforementioned specific processing while using data generation model 58. The data generation model 58 can also be a model finely tuned to output inference results from prompts that do not contain instructions. In this case, the data generation model 58 can output inference results based on prompts that do not contain instructions. The data processing apparatus 12, etc., includes various data generation models 58, including AI other than the generation AI. AI other than the generation AI can be, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and can perform various processes, but is not limited to this example. Furthermore, the AI ​​can also be an AI agent. Furthermore, when the processing of the above-mentioned parts is performed by AI, the processing can be performed partially or entirely by AI, but is not limited to this example. Furthermore, the processing performed by the AI ​​including the generation AI can be replaced by processing in the rule base, and the processing in the rule base can also be replaced by processing performed by the AI ​​including the generation AI.

[0673] Furthermore, the processing of the aforementioned data processing system 10 is performed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the robot 414, but it can also be performed by both the specific processing unit 290 of the data processing device 12 and the control unit 46A of the robot 414. Additionally, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the robot 414 or external devices, and the robot 414 acquires or collects information required for processing from the data processing device 12 or external devices.

[0674] For example, the collection unit is implemented by the control unit 46A of the robot 414 or the specific processing unit 290 of the data processing device 12. For example, the acquisition unit uses the camera 42 or communication I / F 44 of the robot 414 to acquire step data, which is then processed by the specific processing unit 290 of the data processing device 12. For example, the analysis unit is implemented by the specific processing unit 290 of the data processing device 12, which analyzes the data from the collection unit and the acquisition unit. For example, the generation unit is implemented by the specific processing unit 290 of the data processing device 12, which uses a generation AI to generate a menu. For example, the serving unit is implemented by the speaker 240 of the robot 414 and the control object 443 or the specific processing unit 290 of the data processing device 12, which provides the generated menu to the user. The correspondence between each unit and the device or control unit is not limited to the above examples and various changes can be made.

[0675] In the above embodiments, examples of specific processing by the data processing device 12 are given, but the technology disclosed herein is not limited to this, and specific processing may also be performed by the robot 414.

[0676] Furthermore, the emotion-specific model 59, acting as an emotion engine, can determine a user's emotion based on a specific mapping. Specifically, the emotion-specific model 59 can determine a user's emotion based on an emotion graph that serves as a specific mapping (see [reference]). Figure 9 The emotion-specific model 59 can also determine the robot's emotion, and the specific processing unit 290 performs specific processing based on the robot's emotions.

[0677] Figure 9 This is a diagram representing an emotion map 400 that maps multiple emotions. 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 emotion is. On the outer side of the concentric circles, emotions representing states or behaviors arising from mood are arranged. Emotions are concepts that include feelings and mental states. Emotions generated by reactions occurring in the brain are arranged roughly to the left of the concentric circles. Emotions derived from situational judgments are arranged roughly to the right of the concentric circles. Emotions generated by reactions occurring in the brain and derived from situational judgments are arranged roughly above and below the concentric circles. Furthermore, "pleasant" emotions are arranged above the concentric circles, and "unpleasant" emotions are arranged below them. Thus, in the emotion map 400, multiple emotions are mapped based on the structure that generates emotions, and emotions that are likely to occur simultaneously are mapped close to each other.

[0678] These emotions are distributed at the three o'clock position of the emotion map 400, typically fluctuating between peace and anxiety. In the right half of the emotion map 400, situational awareness dominates over internal sensation, thus resulting in an impression of calm.

[0679] The inner side of the emotion map 400 represents the inner state, while the outer side represents behavior. Therefore, the further outward you are from the emotion map 400, the more visible the emotion becomes (manifested in behavior).

[0680] Here, human emotions are based on various balances such as posture and blood sugar levels. When these balances deviate from an ideal state, it indicates an unpleasant state; when they approach the ideal state, it indicates a pleasant state. Emotions in robots, cars, motorcycles, etc., can also be created in the following way: based on various balances such as posture and remaining battery power, when these balances deviate from an ideal state, it indicates an unpleasant state; when they approach the ideal state, it indicates a pleasant state. Emotion maps can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on a Brain Physiological Signal Analysis System for Voice Emotion Recognition and Emotion, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). In the left half of the emotion map, emotions belonging to the sensory-dominated region, called "response," are arranged. Furthermore, in the right half of the emotion map, emotions belonging to the situational cognition-dominated region, called "situation," are arranged.

[0681] In the emotion map, two types of emotions that promote learning are defined. One is a negative emotion on the situational side, in the middle or peripheral region of "repentance" or "reflection." This occurs when the robot experiences negative emotions such as "I don't want to experience this feeling again" or "I don't want to be blamed again." The other is a positive emotion on the response side, near the "desire" region. This occurs when there are positive feelings such as "wanting more" or "wanting to know more."

[0682] The emotion-specific model 59 inputs user input into a pre-trained neural network to obtain emotion values ​​representing each emotion shown in the emotion map 400, thereby determining 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... Figure 10 As shown in the sentiment graph 900, it was trained in a way that sentiments that are configured close to each other have similar values. Figure 10 The text shows examples of emotions such as "peace of mind", "stability", and "reassurance" that have similar emotion values.

[0683] The above description focuses on the functions of the data processing device 12, but the system of this disclosure is not necessarily installed on a server. The system of this disclosure can also be installed as a general information processing system. This disclosure can also be installed, for example, as a software program running on a personal computer, an application running on a smartphone, etc. The method of this disclosure can also be provided to users in the form of SaaS (Software as a Service).

[0684] In the above embodiments, an example of a specific process being performed by a single computer 22 is given. However, the technology disclosed herein is not limited to this, and the specific process can also be distributed among multiple computers, including computer 22. For example, the data generation model 58 can be located on an external device of the data processing apparatus 12, where data is generated based on the input data.

[0685] In the above embodiments, examples of storing a specific processing program 56 in the memory 32 have been described, but the technology disclosed herein is not limited thereto. For example, the specific processing program 56 may also be stored in a portable computer-readable non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed into the computer 22 of the data processing device 12. The processor 28 executes specific processing according to the specific processing program 56.

[0686] Alternatively, a specific processing program 56 may be pre-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 according to the requirements of the data processing device 12.

[0687] In addition, it is not necessary to store all the specific processing program 56 in the storage device such as the server connected to the data processing device 12 via the network 54 or in the memory 32; a portion of the specific processing program 56 may be stored in advance.

[0688] As hardware resources for performing specific processes, various processors, as shown below, can be used. For example, a CPU can be listed as a processor, which functions as a general-purpose processor that performs specific processes by executing software, i.e., a program. Furthermore, processors can be listed as special-purpose circuits such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices), or ASICs (Application-Specific Integrated Circuits), which are processors with circuitry specifically designed to perform specific processes. Each processor has built-in or connected memory, and each processor executes specific processes using that memory.

[0689] The hardware resources for performing a specific process can consist of one of these various processors, or a combination of two or more processors of the same or different types (e.g., a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Alternatively, the hardware resources for performing a specific process can be a single processor.

[0690] As an example of a single processor, there are two approaches: First, a processor is composed of a combination of one or more CPUs and software, which functions as a hardware resource to perform a specific process; second, as represented by a SoC (System-on-a-chip), a processor is used to implement the functionality of the entire system, which includes multiple hardware resources for performing a specific process, using a single IC (Integrated Circuit) chip. In this way, the specific process is implemented by using one or more of the aforementioned processors as hardware resources.

[0691] Furthermore, the hardware architecture of these various processors, more specifically, can utilize circuits that combine semiconductor elements and other circuit components. Moreover, the specific process described above is just one example. Therefore, without departing from the main point, unnecessary steps can certainly be deleted, new steps added, or the processing order changed.

[0692] The descriptions and illustrations above are detailed explanations of a portion of the technology disclosed herein, and are merely one example of the technology disclosed herein. For example, the above descriptions of the structure, function, effect, and results are just one example of the structure, function, effect, and results of a portion of the technology disclosed herein. Therefore, without departing from the spirit of the technology disclosed herein, unnecessary parts may be deleted, new elements added, or replacements may be made to the descriptions and illustrations above. Furthermore, to avoid confusion and facilitate understanding of a portion of the technology disclosed herein, explanations of common technical knowledge that do not require special explanation under the premise of being able to implement the technology disclosed herein have been omitted from the descriptions and illustrations above.

[0693] All documents, patent applications and technical specifications set forth in this specification are incorporated herein by reference to the same extent that each document, patent application and technical specification is specifically and individually described therein and referenced by reference.

[0694] In addition, the following notes are provided in response to the above explanation.

[0695] Example 1 (Note 1) An information processing system, characterized in that it comprises: A device for receiving incoming call signals and obtaining sender identification information through a communication device, and for parsing the incoming call signals to extract the sender identification information; A device for comparing the sender identification information with an information set in an information storage device, and classifying the sender identification information as a known sender or an unknown sender based on whether the sender identification information is included in a pre-registered information set; An apparatus for generating a prompt statement containing the sender identification information and historical information associated with the sender identification information for a sender classified as the unknown sender, inputting the prompt statement into a generative artificial intelligence model to obtain security-related evaluation information, and determining the security of communication from the unknown sender based on the evaluation information. A device for determining whether to allow or deny a communication connection of the communication device based on the security determination result, controlling the communication connection, and recording the security determination result and the sender identification information in the information storage device; An apparatus for sending a notification message containing the sender identification information, the evaluation information, and the security determination result to an information processing terminal when the security determination result is determined to be safe or acceptable, and prompting the user to respond through the display of the notification message to the user on the information processing terminal; An apparatus for receiving connection permission or connection rejection response information from the information processing terminal based on the user response operation, and controlling the communication device according to the response information and the security determination result to perform connection or interruption of voice communication. An apparatus for storing the response information obtained from the user and the security-related evaluation information in the information storage device as learning data to be provided to the generative artificial intelligence model, and for improving the accuracy of subsequent security determination processing for the unknown sender.

[0696] (Note 2) According to the information processing system described in Appendix 1, the notification information includes risk classification and a summary description of the evaluation information generated based on the generative artificial intelligence model, and the information processing terminal provides prompts to the user in different visual display modes according to the risk classification.

[0697] (Note 3) According to the information processing system described in Appendix 1, the incoming call signal reception includes a communication line utilizing optical signals, the system is configured to control an optical signal conversion device for converting between optical signals and electrical signals, acquire the incoming call signal through the optical signal conversion device, and control the optical signal conversion device based on the security determination result and the user response operation.

[0698] Application Example 1 (Note 1) An information processing system, characterized in that it comprises: A device for receiving incoming call signals and extracting communication identification information from the incoming call signals to initiate call processing; A device for comparing the communication identification information with user-related information stored in the information storage unit to determine whether the communication is an unknown sender or a known sender; A device for obtaining historical communication records and user evaluation information for communications that are determined to be from unknown sources, and generating statistical feature information including communication frequency information, call duration information, time period information, and user feedback information. An apparatus for inputting structured information composed of the statistical feature information and prompt statements generated based on the communication identification information into a generative artificial intelligence model and running it, so as to output risk index and risk description information related to the communication. An apparatus for obtaining a numerical risk index based on the output of the generative artificial intelligence model, and determining at least one processing strategy among communication blocking, warning display, call hold, or normal connection by combining pre-stored threshold information and user settings. A device for sending risk level information and processing instruction information to a terminal device according to the processing strategy, causing the display of the terminal device to display warning information or trust information, and controlling the allowance or rejection of call connection through the call connection control unit. A device for receiving connection operation information and number classification information generated by the user from a terminal device, storing the information as call history information and user evaluation information in an information storage unit, and updating it into training data that can be input into the generative artificial intelligence model. An apparatus for periodically extracting call history information and user evaluation information from the information storage unit, performing data processing and statistical processing on them, and training or fine-tuning the parameters of the generative artificial intelligence model accordingly, so as to improve the accuracy of risk assessment for unknown incoming messages. An apparatus for performing evaluation processing on specified communication identification information by the generative artificial intelligence model based on a user's query request, and providing the resulting risk index and explanatory information to a terminal device.

[0699] (Note 2) According to the information processing system described in Appendix 1, the system further includes: a device for performing speech recognition processing on the user's voice input to obtain natural language information, generating appointment processing information or order processing information based on the natural language information, including the appointment processing information or the order processing information in a prompt statement and inputting it into a generative artificial intelligence model to perform communication content generation processing, and generating control information for automatically initiating outbound communication to an external communication network based on the communication content generation result.

[0700] (Note 3) According to the information processing system described in Appendix 1, the system further includes: a communication device for controlling the signal conversion between optical signals and electrical signals; a device for extending the communication device as a communication path control unit to generate control information for controlling the connection mode with the communication network; and a device for switching control of the optical signal path and the electrical signal path according to the control information.

[0701] Example 2 (Note 1) An information processing system, characterized in that it comprises: A device for acquiring voice input and converting the voice emitted by a user into digital voice data via a voice input / output device; A device for performing speech recognition processing on the digital voice data to convert it into text information; A device for generating prompt statements that input the text information into a natural language processing engine including a generative artificial intelligence model, and using the generative artificial intelligence model to parse structured instruction information containing user intent and reservation conditions from the text information; A device for retrieving contact information stored in an information storage device and obtaining contact information of a service provider based on the structured instruction information; A device for automatically initiating calls or data communications through a communication service interface based on the structured instruction information and the contact information, and generating communication control information for making reservations for transportation or facilities; An apparatus for generating a prompt statement for a user-facing notification message based on reservation result information obtained through the communication service interface or service provider device, and sending the notification message to a terminal device to present the reservation result; An apparatus for estimating a user’s emotional state from the text information or the digital voice data, and for changing the prompts or notification messages input to the generative artificial intelligence model based on the emotional state. A device for receiving incoming call signals and extracting caller information, comparing the caller information with the information storage device, performing risk assessment on communications from unknown callers, and generating prompts for security judgment based on the assessment results and inputting them into the generative artificial intelligence model.

[0702] (Note 2) The information processing system according to Appendix 1 is characterized in that, The system is used to parse the text information related to various appointments or orders, generate the structured instruction information using prompts input to the generative artificial intelligence model, and control the communication service interface to automatically initiate communication based on the structured instruction information to generate control information for executing the order processing.

[0703] (Note 3) The information processing system according to Appendix 1 is characterized in that, This system is used to control an optical communication device that receives optical signals, converts the optical signals into electrical signals and connects them to a communication network. It generates results by including control information for communication network connection control in the prompt statements input to the generative artificial intelligence model, and enables the optical communication device to operate in an extended manner based on the control information.

[0704] Application Example 2 (Note 1) An information processing system, characterized in that it comprises: A device for receiving incoming call signals and extracting sender identification information, comparing the sender identification information with stored information to distinguish between known and unknown communications; An apparatus for acquiring the voice of the unknown communication, converting the voice into text information through voice recognition processing, and assessing the risk of sales or fraud based on the text information; A device for blocking or maintaining the unknown communication based on the risk assessment results, and for determining whether to transfer the call to the user; A device for acquiring a user's spoken voice, converting the user's request content into text information through voice recognition processing, generating a prompt statement based on the request content and status information, and inputting the prompt statement into a generative artificial intelligence model to obtain structured instruction information related to the reservation of transportation vehicle or facility use. A device for controlling an external communication interface based on the structured instruction information, and automatically executing outbound call processing for reservation or ordering objects through a communication network; An apparatus for extracting features from a user's voice or image, estimating an emotional state, generating a prompt statement based on the emotional state and the request content, and inputting the prompt statement into a generative artificial intelligence model to obtain response indication information for adjusting response content or service content according to the emotional state. A device for controlling the output of a terminal device based on the response indication information and providing feedback to the user through a display device or an acoustic device; A device for controlling an optical communication device connected to an optical transmission path, converting optical signals into electrical signals to establish a connection with a communication network, and sending and receiving voice data, text information, and input / output data with the generative artificial intelligence model through the communication network. A device for acquiring product images from an image acquisition device, preprocessing the product images and inputting them into a learning model to obtain quality judgment results, and generating control information for controlling the production device to remove defective products based on the quality judgment results. An apparatus for generating prompt statements based on the quality judgment results and production status information, inputting the prompt statements into a generative artificial intelligence model to obtain explanatory information for quality reports, and recording or outputting such information. An apparatus for acquiring user operation records or evaluation information and updating the prompt statements or model operating parameters for the generative artificial intelligence model accordingly, so as to optimize subsequent response content or service content.

[0705] (Note 2) The information processing system according to Appendix 1 is characterized in that, The processor in the system is configured to: parse request content related to various appointments or orders from the user's spoken voice through speech recognition processing, generate prompt statements containing the request content and user attribute information and input them into the generative artificial intelligence model, and automatically control the external communication interface based on the order content and contact information obtained from the generative artificial intelligence model to complete the order processing of goods or services.

[0706] (Note 3) The information processing system according to Appendix 1 is characterized in that, The processor in the system is configured to send and receive control signals and service data with terminal equipment connected to the optical communication device, and generate prompt statements for controlling the timing of input and output data transmission or data compression conditions of generative artificial intelligence models based on the conversion status of optical signals and electrical signals, bandwidth information and time delay information of the communication path, thereby optimizing the operation of the optical communication unit.

Claims

1. An information processing system, characterized in that, include: processor; The processor is configured as follows: Initial processing of incoming calls is performed, and caller information is extracted from the received incoming call signal; Initial processing is performed on communications from unknown callers, and the caller information is compared with a database to perform a risk assessment; Generative artificial intelligence models are used to analyze the relevant information of the caller and generate prompts to determine the security of the caller. The system uses speech recognition technology to parse user requests and generate prompts for AI-initiated communications to handle transportation arrangements and reservations. Identify user emotions and generate prompts based on the user emotion information to provide corresponding services.

2. The information processing system according to claim 1, characterized in that, The processor is also configured to: use speech recognition technology to parse various reservation requests or order requests, and generate prompt information for artificial intelligence to automatically initiate communication to complete the reservation or order.

3. The information processing system according to claim 1, characterized in that, The processor is also configured to: control means for extending to the optical communication unit to convert optical signals into electrical signals and connect them to the communication network, and generate prompts for performing the control and connection operations.

Images