Information processing system

Abstract

This invention provides an information processing system. The information processing system is characterized by comprising: a processor; wherein the processor is configured to: parse input information from a user, generate a prompt indicating the generation of data required for information exchange; input the generated prompt into a generative artificial intelligence model to generate the data required for information exchange; and store the generated data in a recording device and forward it to the user as the recipient.

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 speech in response to the user's speech.

[0003] The problem this invention aims to solve is that existing information exchange systems largely rely on users manually editing and sending messages, lacking intelligent parsing of user input intent and automatic generation of information content. This leads to the following issues: First, users need to repeatedly conceive and edit content when engaging in complex or frequent information exchanges, resulting in low communication efficiency. Second, the system cannot fully utilize generative artificial intelligence models to automatically generate high-quality, structured exchange data based on user input, making it difficult to promptly meet personalized and diverse communication needs. Third, in specific scenarios such as information sharing among multiple users or matchmaking, the system lacks the ability to automatically recommend potential partners and generate communication information tailored to shared interests based on user criteria, thereby reducing the efficiency of establishing effective communication and relationship matching among users. Therefore, there is an urgent need for a system that can automatically generate the data required for information exchange, support information sharing among users, and recommend partners with shared interests and generate corresponding information in environments that meet user-selected criteria, thereby improving the intelligence and efficiency of information exchange. Summary of the Invention

[0004] To address the aforementioned issues, this invention provides an information processing system comprising a processor configured to: parse input information from a user, generate prompts indicating the generation of data required for information exchange; input the generated prompts into a generative artificial intelligence model to generate the data required for information exchange; and store the generated data in a recording device and forward it to the user as the recipient. Through this structure, the system of this invention can automatically generate data conforming to the communication purpose based on user input, and realize the storage and forwarding of the generated data, thereby reducing the burden on users to manually edit content.

[0005] Furthermore, to enable information sharing among users and meet diverse needs, the processor is configured to: parse the information when generating information that can be shared among users, and use prompts to instruct the generation of data according to user needs, thereby generating data that meets user requirements. Thus, the system can automatically generate content suitable for sharing based on the requests of different users, for use in group communication, announcements, or information dissemination within interest groups.

[0006] Furthermore, to improve the efficiency of user matching and communication in specific environments, the processor is configured to: detect recommended users with shared interests in an environment that meets the user's selected criteria, and use prompts to indicate the generation of information to generate information related to the recommended users. By automatically identifying recommended individuals who meet the user's selected criteria and have shared interests in dating or other matching scenarios, and generating targeted communication information for them, the system of the present invention can provide users with conversation entry points and suggestions for communication content, thereby improving the effectiveness of information exchange and user experience.

[0007] "System" refers to an overall apparatus or set of apparatuses including at least one processor and a recording device and / or other hardware and software resources communicatively connected thereto, used to perform the information parsing, data generation, storage and forwarding processing described in this invention.

[0008] A "processor" is a computing unit that can execute program instructions, perform calculations on input data, and make logical judgments. It can be a single physical processor, a combination of multiple processors, or a logical processing unit that is virtually implemented by cloud computing resources.

[0009] "User" refers to the subject that interacts with the system through a terminal device, inputs information, and receives data output by the system. It can be a natural person or an entity with an account identifier in the system.

[0010] "Input information" refers to various types of information submitted by users to the system through terminal devices, including but not limited to text, options, parameter settings, structured or semi-structured data, used to express communication intentions or business needs.

[0011] "Parsing" refers to the process by which a processor analyzes and processes input or existing information, including but not limited to semantic understanding, intent recognition, keyword extraction, structure transformation, and feature extraction, in order to provide a foundation for subsequent data generation.

[0012] "Prompts" refer to control or instruction information generated by the processor and input into the generative artificial intelligence model. They are used to instruct the generative artificial intelligence model to generate text or structured data required for information exchange in accordance with the expected goals, formats, or content requirements.

[0013] "Generative AI models" refer to AI models that can automatically generate corresponding data based on input prompts, including but not limited to large language models, text generation models, and multimodal generation models.

[0014] "Data required for information exchange" refers to the data content that is necessary or useful for enabling information exchange, sharing or interaction between users, including but not limited to text messages, summary descriptions, recommendation information, matching results, dialogue suggestions, etc.

[0015] "Recording device" means a storage device used to store data generated or received by a processor, including but not limited to database servers, hard disks, solid-state storage, cloud storage services or other non-transitory computer-readable storage media.

[0016] "Forwarding" refers to the process by which a processor, after storing the generated data in a recording device, sends the data through a communication network to the target user or their terminal device as the recipient.

[0017] "Information that can be shared among users" refers to data content generated by the system and available for multiple users to view or use together, including but not limited to group announcements, shared documents, mass messages, and public recommendation information.

[0018] "User needs" refers to the expected content, functional requirements, or preferred conditions expressed by users through input information, option settings, or historical behavior during the use of the system.

[0019] "User-selected criteria" refers to a set of conditions that a user pre-sets or specifies during use to filter other users or information, including but not limited to age range, gender, region, interests, values, or other matching indicators.

[0020] "An environment that meets the user's selected conditions" refers to the operating state or data environment in which the system has acquired and is able to utilize information and data related to the user's selected conditions, thereby enabling it to perform filtering, matching, and recommendation processes based on these conditions.

[0021] "Recommended users" refers to other users selected by the system based on predetermined rules or algorithms, provided that the current user meets the selected criteria and has a certain degree of matching with the current user and is suitable for communication with the current user.

[0022] "Shared interests" refers to the fact that the current user and the recommended user have the same or similar interests, topics of interest, or behavioral preferences, such as both liking travel, movies, food, sports, etc.

[0023] "Information related to the recommended user" refers to data content generated by the system for the recommended user to help the current user communicate with that recommended user, including but not limited to opening remarks, topic suggestions, interest descriptions, personalized introductions, or other communication aids. Attached Figure Description

[0024] Figure 1 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the first embodiment.

[0025] 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.

[0026] Figure 3 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the second embodiment.

[0027] 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.

[0028] Figure 5 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the third embodiment.

[0029] Figure 6 This is a conceptual diagram illustrating an example of the main functions of the data processing apparatus and head-mounted terminal according to the third embodiment.

[0030] Figure 7 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the fourth embodiment.

[0031] Figure 8 This is a conceptual diagram illustrating an example of the main functions of the data processing device and robot according to the fourth embodiment.

[0032] Figure 9 This represents an emotion map that maps multiple emotions.

[0033] Figure 10 This represents an emotion map that maps multiple emotions.

[0034] Figure 11 This is a sequence diagram illustrating the processing flow of the data processing system of the first embodiment.

[0035] Figure 12 This is a sequence diagram illustrating the processing flow of the data processing system in Application Example 1.

[0036] Figure 13 This is a sequence diagram illustrating the processing flow of the data processing system of the second embodiment.

[0037] Figure 14 This is a sequence diagram illustrating the processing flow of the data processing system in Application Example 2. Detailed Implementation

[0038] Hereinafter, an example of an implementation of the system according to the present disclosure will be described with reference to the accompanying drawings.

[0039] First, let me explain the terminology used in the following instructions.

[0040] 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.

[0041] 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.

[0042] 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.

[0043] 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.

[0044] 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.

[0045] First Implementation Method Figure 1 An example of the configuration of the data processing system 10 according to the first embodiment is shown.

[0046] 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.

[0047] 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).

[0048] 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.

[0049] 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.

[0050] 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.

[0051] 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.

[0052] Figure 2 The diagram shows an example of the main functions of the data processing device 12 and the smart device 14.

[0053] 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.

[0054] 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).

[0055] 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.

[0056] 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.

[0057] 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."

[0058] In existing information exchange systems, servers typically act as passive forwarding nodes, simply receiving, storing, and forwarding message data from terminals. Such systems suffer from the following technical problems: First, servers only perform format conversion and routing control on user input, lacking semantic parsing and abstract modeling of the input information. This makes it difficult for the system to automatically generate high-quality exchange content based on complex and multidimensional user needs. Second, the interaction between the server and the generative AI model is often a simple "question-answer call," lacking a mechanism for dynamically constructing prompts based on communication scenarios and historical data. This prevents the full utilization of the generative AI model's generation capabilities to optimize message content and interaction processes. Third, the server's online / offline state management and message persistence strategies are often designed independently, lacking tight coupling with generated content. This results in inconsistent message delivery under different states and difficulty in unified management of message history, affecting system reliability and scalability. Fourth, the processing of multi-user shared content and user discovery recommendations is often implemented through independent business logic or external services, without integrated design with the underlying message storage, forwarding control, and generative AI model calls. This leads to low resource utilization efficiency and complex system maintenance.

[0059] Therefore, a technical solution is needed that integrates structured processing of user input information on the server side, generation of prompt statements, invocation of generative artificial intelligence models, unified storage control, and online state-aware forwarding control. This solution aims to improve the overall architecture and processing flow of information exchange systems at the computer level, and to automate, intelligentize, and improve the performance of message generation, storage, and distribution. This will enhance the processing efficiency, scalability, and service quality of computer systems in complex communication scenarios.

[0060] 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.

[0061] In this invention, the server includes: a data processing device for receiving input information from a user terminal via a communication link and converting the input information into intermediate data representing structured exchange information; a generation device for generating a prompt statement based on the intermediate data and using the prompt statement to instruct a generative artificial intelligence model to generate exchange information for information exchange and / or a method for generating the exchange information; an acquisition device for inputting the prompt statement into the generative artificial intelligence model and acquiring response data including the exchange information and / or the generation method; and a device for converting the response data into a write request to a recording device and controlling the recording device to persist the response data. The system includes: a storage control device for centralized storage; a forwarding control device for sending the response data instantly via a real-time communication link when the target user is online and when the target user reconnects when offline, based on the target user's identification information and communication status information corresponding to the response data; a history management device for associating the source user input information, source user identification information, and terminal identification information corresponding to the response data and managing them uniformly in the form of message history in a recording device; and a status notification device for feeding back status information, including the response data storage result and the forwarding result to the target user, to the source user terminal so that the source user terminal can perform display control based on the status information. This allows for the formation of an end-to-end processing link within the server, encompassing "user input parsing—prompt statement construction—generative artificial intelligence model invocation—unified persistent storage—online status-aware forwarding—historical consistency management." This optimizes data structures and processing flows at the computer level, enabling efficient processing of complex session scenarios, improving system throughput and scalability, and enhancing information exchange quality and user experience through intelligent generation and unified control mechanisms.

[0062] A "system" refers to a collection of devices consisting of multiple information processing components that work together through communication to achieve functions such as information reception, processing, storage, and forwarding.

[0063] "User terminal" refers to a computing device operated by a user for inputting, sending, receiving and displaying information, including but not limited to mobile terminals, fixed terminals and other electronic devices with network communication functions.

[0064] A "communication link" refers to a physical or logical connection used to transmit data between a user terminal and a server, including various network connection paths such as wired and wireless communication.

[0065] "Input information" refers to the raw data content submitted by the user through the user terminal and sent to the server via the communication link, including text data and identification and control information related to the text.

[0066] "Intermediate data" refers to a data structure generated by the server based on input information, used to represent the exchanged information content and related metadata in a structured form. This data structure is suitable for use by subsequent processing modules and generative artificial intelligence models.

[0067] "Exchange information" refers to content data used for information exchange between different users, including text messages and the identification information, time information and context information associated with those text messages.

[0068] "Generative artificial intelligence models" refer to intelligent computing models that are trained on large amounts of data and can automatically generate text and other content based on prompts. They can generate new exchange information or related explanatory data based on the input context and instructions.

[0069] "Prompt statements" refer to instructional texts constructed by the server and input into the generative AI model to instruct the generative AI model to perform specific generative tasks, including generating exchange information, abstract explanatory data, introductory messages, or session initiation messages.

[0070] "Response data" refers to the data set containing generated content output by the generative artificial intelligence model after receiving a prompt statement, including the generated exchange information, the description of the generation method, and related metadata.

[0071] "Recording device" refers to an information storage component used to persistently store response data and its associated information, including databases, storage servers, or other non-volatile storage media.

[0072] "Data processing device" refers to the processing unit within the server that receives data sent by user terminals and parses, formats, and structures it. It can generate intermediate data from the input information through program execution.

[0073] "Generation device" refers to the processing unit within the server that automatically constructs prompt statements based on intermediate data and instructs and constrains the generation behavior of generative artificial intelligence models.

[0074] "Acquisition device" refers to the interactive unit within the server used to send prompts to the generative artificial intelligence model and receive its output response data.

[0075] "Storage control unit" refers to the data management unit within the server that converts response data into write requests for the recording device and controls the persistent storage of the response data.

[0076] "Forwarding control device" refers to a data forwarding unit within a server that controls whether response data is sent immediately when online or delayed when offline, based on the target user's identification information and communication status information.

[0077] "Historical management device" refers to a management unit within the server that associates the source user's input information, source user identification information, terminal identification information, and response data, and manages them uniformly in the form of message history in the recording device.

[0078] "Status notification device" refers to a notification unit within the server used to generate and send status information to the source user terminal to indicate the storage results of the response data and the forwarding results to the target user.

[0079] "Target user" refers to a user entity that is designated as the recipient of response data or exchanged information in the system.

[0080] "Source user" refers to the user entity that initiates input information in the system and triggers the generation or forwarding of response data.

[0081] "Identification information" refers to symbolic data used to uniquely or relatively uniquely distinguish users, terminals, or session objects within the system, including user identifiers, terminal identifiers, and session identifiers.

[0082] "Communication status information" refers to data used to indicate the current connection status between the target user and the server, including status indicators such as online, offline, and reconnection.

[0083] "Historical information" refers to the data set stored in the recording device about past information exchange processes, including past input information, response data, forwarding records and related metadata.

[0084] "Historical format" refers to a data structure that organizes and displays messages and their associated information in chronological order or session dimension, and is used to reflect the temporal and contextual relationships of information exchange.

[0085] "Common concerns" refers to topics or themes that are extracted from input and historical information from multiple users and that have common interests or needs among multiple users.

[0086] "Abstract explanatory data" refers to generated content that describes a matter of common concern in a general and inductive manner, and is intended for sharing or recommendation to multiple users.

[0087] "Behavioral history information" refers to data generated based on user behavior and stored by the recording device during system operation, including message sending records, receiving records, reading records, and related interaction behavior records.

[0088] "Recommended user group" refers to a set of users that are similar or related in terms of interests, attributes or behavioral patterns, which are automatically detected by analyzing historical behavioral information and selected conditions.

[0089] "Introduction messages" refer to generated content used to establish connections between recommended user groups, introduce each other's information, and promote further communication.

[0090] "Session initiation message" refers to the generated content used to initiate a new conversation between two or more users and guide the start of communication.

[0091] In one embodiment of the present invention, the server is implemented on a computer hardware platform, which includes a processor, main memory, non-volatile memory, a network interface, and a storage device connected thereto. The server runs application server software, network server software, and database management software on an operating system (e.g., a Linux-based server operating system). The server communicates with multiple terminals through the network interface and performs data read / write operations with a recording device via an internal bus.

[0092] The server's software architecture includes logical modules such as data processing devices, generation devices, acquisition devices, storage control devices, forwarding control devices, history management devices, and status notification devices. These modules can be deployed on the same physical server or multiple distributed servers as processes, threads, or function libraries. The server maintains a relational database in its storage device, which can be implemented by a database management system, such as relational database management software. The server can also further utilize in-memory databases or caching systems to store user online status information and session index information.

[0093] The terminal can be a mobile electronic device, a fixed electronic device, or other computing device with display and input functions, typically including smartphones, tablet computers, and personal computing devices. The terminal includes a processor, memory, a touchscreen or keyboard, a display screen, an audio output module, and a wireless or wired communication module. The terminal runs a messaging application on its operating system (e.g., a mobile terminal operating system or a desktop operating system), which provides a user interface for inputting text, displaying message content, and communicating with a server over a network.

[0094] Users operate the messaging application through a terminal. Users input natural language text into the interface, such as typing message text in a conversation window with another user or a smart assistant, and trigger the sending action through touch or click. After capturing the input content and user actions, the terminal performs local preprocessing on the input data, including character encoding conversion, sensitive character filtering, and local length checking. The terminal encapsulates this input information, along with user identifier, session identifier, time information, and terminal identifier, into a structured object and generates a network request using a network library.

[0095] The terminal establishes a secure communication connection to the server via its network communication module (e.g., through HTTPS or an encrypted tunnel) and sends structured objects to the server using application layer protocols. The terminal maintains the message sending status locally. Upon receiving status information from the server, the terminal updates its local interface display accordingly, such as changing the message bubble from "sending" to "delivered," or displaying an error indicator if sending fails. When receiving response data pushed by the server, the terminal parses the structured data using its local parsing module, inserts the message content into the interface message list, and, if necessary, calls the operating system notification interface to emit sound or vibration alerts to the user.

[0096] After receiving a network request from a terminal, the server uses network server software to parse and route the request, then passes the request content to the application module. The server performs syntax- and semantic-level parsing of the input information in its data processing unit, constructing intermediate data structures. The server can use object-oriented data structures or key-value mapping structures to represent the intermediate data, which includes: user identifier, session identifier, original message text, language, timestamp, terminal identifier, context window index, and control parameters related to subsequent generative artificial intelligence model calls.

[0097] The server constructs prompt statements based on intermediate data in the generation device. These prompt statements are not simply forwarded directly to the generative AI model; instead, the server generates them by combining rules based on the session context, historical messages, and system policies. The server can pre-maintain a set of prompt templates in storage, each template containing system role descriptions, constraints on generation style, output length limits, and security filtering rules. In the generation device, the server matches an appropriate template based on the intermediate data, inserts the user's current input, relevant historical fragments, and system commands into the template, forming a complete prompt statement.

[0098] For example, when a user requests an explanation of how to use the system, the server can generate the following prompt: "System: You are an assistant that helps users understand how to use the chat application."

[0099] User: Please tell me how to effectively use this chat application to stay connected with friends. For example, when the server needs to generate technical documentation for developers, it can generate the following prompt: "Based on the following system design, please explain in detail how user A uses a mobile chat application to send a text message to user B, and explain how the terminal constructs an HTTP request, how the server parses the request, and how it saves and forwards the message through the database." The server can add explicit structured instructions to the prompts in the generation device, such as requiring the generative AI model to output step-by-step instructions, restricting the use of specific terminology, and avoiding sensitive content. The server can also strategically truncate or compress historical information based on the length of the current session and the size of the model's context window. This non-conventional processing flow based on context window management can reduce the transmission and processing of invalid data while ensuring generation quality, thereby reducing communication load and computational burden.

[0100] The server interacts with the generative AI model within the acquisition device. In this embodiment, the generative AI model can be deployed locally on the server or provided by a remote service. When deployed locally, the server can employ a deep neural network-based language model architecture, such as a transformer network with multi-layer self-attention mechanisms. This model includes word embedding layers, multi-head self-attention layers, feedforward network layers, and normalization layers. The parameters of each layer are optimized during training using gradient descent. During training, the server uses a large-scale text corpus, defining the prediction of the next word or a fill-in-the-blank task as the learning objective. It employs a cross-entropy loss function as the error function, calculates gradients using backpropagation, and updates weight parameters using an adaptive learning rate optimization algorithm. The server can fine-tune the pre-trained model using domain-specific dialogue data to obtain generative behaviors more suitable for information exchange scenarios.

[0101] During the model inference phase, the acquisition device converts the prompt statements into a labeled sequence, generates feature vectors based on the vocabulary mapping, and feeds them into the generative artificial intelligence model. The model encodes the correlations between different words within the prompt statements in a multi-layered self-attention structure, dynamically aggregating contextual information using attention weights to generate new hidden representations. In the output layer, the server generates a response text sequence based on the hidden representations using probability distribution sampling or decoding algorithms (e.g., greedy decoding or bundle search). The server uses deduplication and post-processing rules in the acquisition device to standardize the generated text, such as unifying punctuation, replacing prohibited words, and controlling paragraph structure.

[0102] When designing and using generative AI models, the server does not employ traditional rule matching or template replacement methods. Instead, it generates new content by inputting structured prompts into the model and utilizing its large-scale internal parameters and statistical relationships learned during training. The high-dimensional vector space and attention weight distribution built within the model enable it to abstractly process user input and context based on patterns learned from a vast corpus. This approach differs from human editing based on explicit rules. The server uses techniques such as configuring temperature parameters, maximum generation length, and coefficients to penalize repetition to achieve diversity control and redundancy suppression, thereby achieving an adjustable balance between fluency and accuracy of the generated content within the computer.

[0103] The server converts the response data into database write operations within the storage control device. The server constructs record rows in a relational database based on the user identifier, session identifier, and time information in the response data. The database may include message tables, user tables, session tables, and status tables. The message table contains message identifiers, source user identifiers, target user identifiers, session identifiers, message text, creation time, status, and metadata fields related to model generation. The server generates structured query statements using a database driver and persists the response data to the recording device. In some implementations, the server may also record summaries of prompts used by the generative AI model and model version identifiers for subsequent analysis and model quality tracking.

[0104] The server utilizes an in-memory data structure in its forwarding control mechanism to maintain user online status and connection session information. The server can maintain a record for each online user, including the user identifier, the session identifier of the current connection, the connection type (e.g., long connection, push channel), the last heartbeat time, and network address information. The server periodically updates this structure based on heartbeats or connection events. When determining if a new message needs to be forwarded, the server checks this structure to determine if the target user is currently online. If online, the server immediately pushes the response data to the target terminal via a real-time communication channel (e.g., long connection or bidirectional communication channel); if offline, the server marks the message status as undelivered in the database and resends it by batch retrieving undelivered message records when the user re-establishes the connection. This online status awareness and forwarding strategy, combined with unified database management, ensures message order consistency and delivery reliability under different network conditions, technically improving the fault tolerance of distributed communication systems under weak network conditions.

[0105] The server associates the source user's input information, source user identifier, terminal identifier, and response data in the historical management device. By establishing indexes and foreign key relationships in the database, the server enables efficient queries by session, user, and time. The server can use an inverted index structure to index message text with keywords, thereby quickly locating message records when relevant historical content needs to be retrieved. This structured storage and indexing method offers higher query efficiency and scalability compared to simple log file storage, helping to maintain system responsiveness in scenarios with a large number of users and massive amounts of messages.

[0106] The server generates status information for each write and forwarding operation in the status notification device. The server encapsulates the database write results (success or failure, number of records, message identifier) ​​and forwarding results (delivered, pending, reason for failure) into structured status data and sends it to the source user terminal via response or asynchronous push. The terminal updates its interface display based on the fields in the status data, such as displaying a delivered indicator, a retry button, or error message text. Through this fine-grained status notification mechanism, the system can automatically guide the terminal to retry, cache, or display prompts without increasing manual user intervention, thereby improving human-computer interaction efficiency.

[0107] In a further embodiment, the server uses a generative artificial intelligence model to abstract multi-user historical information, generating abstract descriptive data that can be shared by multiple users. The server analyzes the content of multiple sessions in the recording device through a data processing device, extracts common concerns, and instructs the model to output concise summaries through specific prompts. For example, the server can generate the following prompt: "Based on the recent chat logs of the following users, summarize the three main topics they are concerned about and describe them in the form of brief entries." After obtaining the summary output by the model, the server stores it as shared exchange information and distributes it to relevant users. This processing method, which uses deep neural networks to abstract high-dimensional dialogue data, can achieve topic clustering and summarization algorithmically within the server without relying on manual rule maintenance, thereby improving the efficiency of information processing under large-scale data.

[0108] In another implementation, the server detects and recommends user groups based on selected criteria and behavioral history information. The server uses feature vectors to represent user behavior in its data processing unit, constructing high-dimensional features such as message frequency, keyword distribution, and session engagement. The server can employ clustering algorithms (e.g., K-means, spectral clustering) or similarity measures (e.g., cosine similarity) to aggregate and analyze user feature vectors, thereby detecting user groups with shared interests or similar attributes. The server then generates prompts, instructing the generative artificial intelligence model to construct introductory messages or session initiation messages, such as: "Please generate a polite introductory message for a group of users with similar interests in technical topics, encouraging them to start a conversation and briefly explaining the specific topics that can be discussed." The server stores the generated introductory message as a special type of exchange information in a recording device and sends it to at least some members of the recommended user group through a forwarding control device. This process combines vector representation, clustering algorithms, and generative artificial intelligence models, enabling the user discovery and matching process to have a high degree of automation and statistical robustness. Compared with traditional recommendation systems based on manual rules, it has better scalability and accuracy in complex behavioral patterns and high-dimensional feature spaces.

[0109] In terms of technical effectiveness, the system of this invention does not merely automate the human action of writing messages, but rather optimizes the representation, processing flow, and storage structure of text and conversation data within the computer. Through intermediate data structures and prompt generation strategies, the server maps user input to a representation space that the model can efficiently process, achieving context-sensitive generation. Through the multi-layered self-attention structure and parameter learning methods of deep neural networks, the server utilizes a high-dimensional feature space to capture complex language patterns, improving the coherence and relevance of generated content. Through unified persistent storage and online state-aware forwarding control, the server ensures message consistency and reachability in a distributed network environment. Through historical management and index structure design, the server maintains good retrieval and statistical performance in scenarios with massive amounts of messages.

[0110] Therefore, this invention improves the processing speed, generation quality, data management capabilities, and communication efficiency of the information exchange system at both the hardware and software levels through the coordinated operation of modular data processing devices, generation devices, acquisition devices, storage control devices, forwarding control devices, and historical management devices. It is an improvement on computer technology itself, rather than being limited to the automation of business processes.

[0111] use Figure 11 The processing procedure is explained.

[0112] Step 1: The user enters a message on the terminal and triggers the send operation.

[0113] Input: The user's natural language text (e.g., "Hi, how's work going lately?"), the current session identifier, and the recipient identifier.

[0114] Output: The message object to be sent within the terminal.

[0115] In the terminal's messaging application interface, the user enters text content in the input box via touchscreen or keyboard and clicks or taps the "Send" button. After capturing the click event, the terminal reads the text string from the input box and combines this string with the user identifier, target user identifier, session identifier, current timestamp, and terminal identifier to construct a message object in memory. The terminal performs basic checks on this object (e.g., checking that the content is not empty and the length does not exceed the limit). If the checks pass, the terminal writes it to the local sending queue and prepares it for the network sending process.

[0116] Step 2: The terminal converts the message object to be sent into a network request and sends it to the server.

[0117] Input: The message object to be sent within the terminal.

[0118] Output: The network request sent to the server and the corresponding sending status flag.

[0119] The terminal calls the network library provided by the operating system to serialize the message object into a key-value pair data structure, and then encodes it into JSON text. The terminal sets the target address of the network request to the server's message receiving interface, uses the JSON text as the request body, and adds header information (including content type and authentication token). The terminal sends the request through a network module (such as a Wi-Fi or mobile communication module), marks the message status as "sending" locally, and indicates this on the interface with a gray bubble or icon, while simultaneously initiating an asynchronous callback to wait for the server's response.

[0120] Step 3: The server receives network requests sent by the terminal and parses them into intermediate data.

[0121] Input: A network request from the terminal (containing message data in JSON format).

[0122] Output: The intermediate data structure inside the server.

[0123] The server receives the request through a listening port in its web server software, parses the HTTP headers and request body, and extracts the JSON text. In its data processing unit, the server calls a JSON parsing library to parse the JSON text into internal data structures (such as objects or key-value maps), obtaining a set containing the source user identifier, target user identifier, session identifier, message text, timestamp, and terminal identifier. Based on this, the server performs syntax-level validation (field existence, type correctness) and authorization validation (authentication token validity), and packages the validated data into intermediate data in a unified format for subsequent prompt generation and storage.

[0124] Step 4: The server generates prompts for invoking generative artificial intelligence models based on intermediate data and historical information.

[0125] Input: The server's internal intermediate data structure and some historical message records related to this session.

[0126] Output: The prompt text constructed for the current task.

[0127] In the generation device, the server first reads the most recent historical messages from the recording device based on the session identifier. The server then constructs the session context according to a predefined strategy (e.g., selecting the N most recent messages in chronological order). Next, the server selects a prompt template from the template library that corresponds to the current task type, such as templates for explanations, general responses, or summaries. The server inserts the current user input text and the selected historical messages into placeholder positions in the template, adding system role descriptions and output constraints to construct the complete prompt statement. During this process, the server may compress or summarize the historical text to meet model context length constraints, thereby reducing redundant content while preserving key information.

[0128] Step 5: The server inputs the prompt statement into the generative artificial intelligence model and obtains the response data.

[0129] Input: The prompt text generated by the server and the model configuration parameters (such as maximum output length and temperature coefficient).

[0130] Output: A response data object containing the generated text.

[0131] The server segments or encodes the prompt statement in the acquisition device, converts it into a token sequence using a vocabulary, and constructs an input tensor according to the model requirements. The server then feeds this tensor, along with configuration parameters, into the generative artificial intelligence model. A multi-layered self-attention network within the model performs forward propagation computation on the token sequence, calculating the hidden representation at each layer and generating a probability distribution for subsequent words at the output layer. On the output side, the server uses a decoding algorithm (e.g., greedy or bundle search) to progressively select or sample the next word based on the probability distribution until a termination condition is met. The server maps the generated token sequence back to a text string to form the response text, and combines the response text with the source user identifier, target user identifier, session identifier, etc., to form a response data object for subsequent storage and forwarding.

[0132] Step 6: The server persistently stores the response data in a recording device.

[0133] Input: Response data object (including generated text and related identification information).

[0134] Output: Persistent message records saved to the recording device, and updated response data objects with database identifiers.

[0135] The server generates database records in the storage control unit based on the response data, constructing a set of fields including source user identifier, target user identifier, session identifier, generated text, creation time, message status, and model metadata. The server generates an insert statement through the database driver and sends it to the relational database. After completing the write, the database returns the insert result and the primary key identifier of the new record. The server adds this primary key identifier to the response data object, updates the message status to "unread" or "pending delivery," and records the message in the index structure for subsequent retrieval and statistics.

[0136] Step 7: The server determines how to forward response data based on the target user's communication status.

[0137] Input: Response data object with database identifier, and online status table in memory.

[0138] Output: Real-time push messages sent to the target terminal or status updates recorded as messages awaiting delivery.

[0139] The server reads the target user identifier from the forwarding control device and queries the corresponding record in the online status table. If the online status table indicates that the user currently has a valid connection, the server constructs a message structure suitable for the real-time channel, encapsulates the response text and necessary metadata into push data, and sends it to the target terminal through the real-time channel interface. If the online status table indicates that the user is offline or has no valid connection, the server only maintains an "undelivered" status in the database and does not send the message immediately. The server can add the identifier of the undelivered message to the offline queue so that it can perform batch push when the user connects again. This branching process based on online status enables the system to reduce unnecessary network transmission while ensuring message reliability.

[0140] Step 8: The terminal receives the response data pushed by the server and displays it on the interface.

[0141] Input: Real-time push data or offline synchronized data from the server (including generated text and metadata).

[0142] Output: Newly added message items in the terminal's local message list and their corresponding interface display.

[0143] When the terminal receives push data on the established real-time connection, it parses the structured data through the parsing module, extracting the message text, sender identifier, session identifier, and time information. The terminal inserts the message into a local message storage structure (such as a memory list or local database) and selects or creates the corresponding dialog window based on the session identifier. If the user currently has the session open, the terminal calls the UI rendering engine to add a message bubble to the message list and adjusts its arrangement and scrolling position according to local style rules. If the user does not have the session open or the application is in the background, the terminal generates a system notification through the notification interface, with a partial text preview attached, to prompt the user to view it.

[0144] Step 9: The server sends storage and forwarding status information back to the source user terminal, and the terminal updates the message status display.

[0145] Inputs: Database write results, forwarding execution results, and session information of the source user terminal.

[0146] Output: Status notification data sent to the source user terminal and the updated interface status.

[0147] The server constructs a status information object in the status notification device based on the return value of the database write operation and the forwarding control result. This object includes a message identifier, whether the write was successful, whether it has been forwarded to the target user, and, if it failed, an error code and reason. The server sends this status information to the source user terminal via a response or a separate push channel. Upon receiving the status information, the terminal parses the message identifier, looks up the corresponding message in its local message list, updates its status from "sending" to "delivered" or "failed to send," and displays a prompt or provides a retry button to the user if the message fails. Through this round-trip process, the terminal can automatically adjust its display and subsequent operations based on the objective status information provided by the server, thereby reducing the burden of manual confirmation and resending by the user.

[0148] 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".

[0149] In existing computer-based information exchange systems, responses or recommendations are typically generated solely based on the user's current text input, using preset rules or simple recommendation algorithms. These systems suffer from the following technical problems: First, servers lack the comprehensive modeling capability for the overall conversation context and long-term user behavior data. Processing only single messages makes it difficult to detect conversation pauses or user lack of topic clues in a timely and accurate manner, thus failing to provide targeted conversation topics at appropriate times. Second, while servers can invoke generative AI models to generate text, they often only use the current message as direct input, failing to structurally embed the user's historical usage records, interest attributes, and conversation history into the prompts. This results in low relevance between the generated results and user preferences and the current conversation context, leading to unstable data quality. Third, when processing user-shared link information, most systems simply forward or display the original link without using a unified data processing workflow to automatically parse and extract metadata from the link resources. This prevents the structured association of external resource titles, summaries, and images with the conversation content and recommendation results, increasing the burden on users to understand the information. Fourth, the server's storage and reuse mechanism for generated results and link metadata is relatively weak, failing to form a unified data structure that can be reused for subsequent session analysis and model suggestion construction, thus limiting the continuous optimization space of recommendation algorithms and natural language processing modules.

[0150] Therefore, it is necessary to provide an improved computer implementation system. By integrating session state detection, prompt statement construction based on user interest attributes, generative artificial intelligence model invocation, and link metadata extraction and unified storage on the server side, the system can improve data processing and data operation methods within the system. This will enhance the relevance, timeliness, and usability of the generated session auxiliary information and recommendation information, thereby achieving an overall improvement in computer information processing capabilities and human-computer interaction efficiency.

[0151] 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.

[0152] In this invention, the server includes means for parsing input information from a user terminal and generating prompt statements for indicating the generation of conversation topics or recommendation information based on the input information and the user's usage history; means for inputting the prompt statements into a generative artificial intelligence model to generate data for information exchange; means for retrieving and parsing the user's past usage history from a storage device to extract interest attributes and reflect the interest attributes in the prompt statements; means for parsing user messages and dialogue history through natural language processing to detect when a conversation is stagnant or lacks topic clues; means for filtering and shaping the data for information exchange according to interest attributes and conversation status into candidate information for prompting the user; means for parsing link information contained in user-shared information, retrieving metadata from external resources and converting it into a form that can be displayed on the user terminal; and means for storing the generated data for information exchange and the metadata in a storage device and forwarding it to the target user terminal. This allows for the automatic detection of session stagnation within the server through an integrated data processing workflow. It dynamically constructs prompts containing user interest attributes and conversation history, efficiently invokes generative artificial intelligence models to generate conversation topics and recommendation information that are highly relevant to the context, and unifies the modeling and storage of metadata from external linked resources with the generated results. This improves the natural language processing capabilities, recommendation accuracy, and the ability to parse and display external resources in information exchange systems at the computer technology level, enabling more efficient and intelligent human-computer interaction.

[0153] A "system" refers to a computer implementation consisting of multiple information processing components such as servers, user terminals, and storage devices that work together through a communication network to perform processes such as information parsing, data generation, data storage, and data transmission.

[0154] A "server" is an information processing device equipped with a processor and storage resources, which communicates with user terminals via a network to perform functions such as input information parsing, prompt generation, calling generative artificial intelligence models, metadata extraction, and data transfer.

[0155] "User terminal" refers to an electronic device operated by a user and used to send and receive data with a server via a network, including but not limited to smartphones, tablets, personal computing devices, or other terminal equipment with communication and display functions.

[0156] "Input information" refers to the content entered by the user on the user terminal and sent to the server via the communication network for information exchange, including text messages, additional link information, and other structured or unstructured data related to the session.

[0157] "Usage history" refers to the collection of behavioral data recorded by the server based on a user's past operations in the system, including historical conversation content, click records, favorite records, browsing records, and records of interactions with recommended information.

[0158] "Past usage history" refers to user usage history data that has been stored by the server before the current session or the current request, used to characterize the user's behavioral characteristics over a long period of time or a certain period of time.

[0159] "Interest attributes" refer to abstract features generated by the server after statistically analyzing the user's usage history to characterize the user's tendencies in content categories, topic preferences, or interaction patterns. These features include interest tags, weight parameters, classification results, and other information.

[0160] "Prompt statements" refer to text or structured instructions constructed by the server and provided to the generative AI model. These instructions are used to guide the generative AI model to generate conversation topics or recommended information based on the specified context, user interest attributes, and the current task.

[0161] "Generative AI models" refer to AI models that can automatically generate natural language text or other data content based on input prompts. They are generally based on deep learning structures (such as neural networks) and can be used for tasks such as dialogue generation, content recommendation description generation, or topic generation.

[0162] "Information exchange data" refers to content data generated by generative artificial intelligence models based on prompts and processed by servers for use in information exchange between users. This includes conversation topics, recommended information, explanatory text, and related auxiliary information.

[0163] "Conversation topics" refer to text content used to guide or sustain conversations. These are natural language expressions generated by generative artificial intelligence models and used as openings, extensions, or transitions in conversations between users.

[0164] "Recommended information" refers to content generated or selected based on user interests and the current context to provide users with personalized suggestions or references, including but not limited to work titles, project lists, article links and their descriptions.

[0165] "Conversation history" refers to a sequence of multiple messages sent continuously by multiple users in the same session and recorded by the server, used to reflect the conversation context, semantic structure, and interaction process.

[0166] Natural Language Processing (NLP) refers to the technical process by which servers perform word segmentation, part-of-speech tagging, semantic analysis, classification, and other processing on user messages, dialogue history, and other text data to extract structured features and semantic information from natural language.

[0167] "The session is in a stagnant state" refers to a session state in which the server, based on the analysis of the dialogue history and user input characteristics, determines that the current session is difficult to continue naturally, lacks new interactive content, or has a clear tendency to become quiet.

[0168] "Lack of topic cues in the conversation" refers to a state in which the server determines, based on user input or behavioral patterns, that the user is having difficulty finding a suitable topic to communicate with. This includes situations where the user explicitly expresses "nothing to talk about" or implicitly indicates a need for topic guidance.

[0169] "Candidate information" refers to a set of alternative content that the server filters, shapes, and sorts the generated conversation topics or recommendation information for display to the user for selection or direct use.

[0170] "Link information" refers to network resource identifiers that appear in user input information, including Uniform Resource Locators (URLs) and other address data that point to external web pages, files, or services.

[0171] "External resources" refers to network resources that are linked to by the information and located outside this system, including web page content, data returned by application programming interfaces, media files, and other data entities that can be accessed via the network.

[0172] Metadata refers to the additional data that a server extracts from external resources to describe the content of those external resources, including title information, summary information, image information, and other descriptive information used to represent the characteristics of the resource.

[0173] "Title information" refers to text data extracted from external resources that briefly identifies the subject or name of the resource.

[0174] "Summary information" refers to a short text description extracted from external resources that summarizes the main content or key points of the resource.

[0175] "Image information" refers to identification or linking information of visual data associated with external resources and used for preview or illustration, including thumbnail addresses or reference data of representative images.

[0176] "Storage device" refers to a computer-readable storage medium used to persistently store user usage history, generated information exchange data and metadata, including magnetic storage media, solid-state storage media or other non-volatile storage resources.

[0177] In a preferred embodiment, the server is deployed on a computer cluster in a data center, which includes computing devices equipped with multi-core central processing units and graphics processing units. The server can run a software environment based on a general-purpose operating system, such as a UNIX-based server operating system. The server can use reverse proxy software and web server software to provide an application programming interface (API) based on the HTTPS protocol. Internally, the server can employ application development frameworks to implement business logic, such as web frameworks based on interpreted languages ​​or service frameworks based on scripting languages. The server can manage structured or semi-structured data such as user history, conversation history, generated results, and link metadata through a database management system (e.g., a relational database management system or a document-oriented database management system).

[0178] In this implementation, the terminal is an electronic device operated by the user. The terminal can be a smart terminal running a mobile operating system, a portable terminal, or a personal terminal running a desktop operating system. The terminal can install client applications, which can be implemented based on a mobile development framework or a general user interface framework. The terminal can communicate bidirectionally with the server via Transmission Control Protocol / Internet Protocol (TCP / IP) stacks and a secure version of Hypertext Transfer Protocol (HTTP). The terminal can utilize a local display module and an input module to accept user input and display information output by the server.

[0179] In this implementation, users input text messages, select recommended topics, or click links through the terminal's graphical user interface. When users input text on the terminal interface, the terminal encodes the input content into a byte sequence in a unified character encoding format, adds a user identifier, session identifier, and timestamp to form a data structure, and then sends it to the server via the network protocol stack. Users do not need to understand the internal processing details of the server to enjoy conversation assistance and recommendation services through natural language interaction.

[0180] In one implementation, the server uses natural language processing software components to parse the input information from the terminal. The server can employ a pre-trained semantic model based on a Transformer architecture, such as a bidirectional encoding representation model or equivalent structure, to lexically and vectorize the message text. The server can use a tokenizer to split the input sentence into sub-word units and map each sub-word to an integer index. The server can then input the index sequence into a deep neural network to obtain a sentence-level vector representation. The server can combine these vectors with features such as time interval, message length, and message type label to construct a feature vector representing the current session state.

[0181] In this implementation, the server can use a machine learning classifier (e.g., logistic regression, gradient boosting tree, or shallow neural network) to determine whether a conversation is stalled. During offline training, the server can perform supervised learning using a large amount of labeled dialogue datasets, separately labeling instances of "natural conversation continuation" and "stagnant or lacking topics," and updating model parameters using cross-entropy loss, stochastic gradient descent, or variations thereof. During inference, the server receives the feature vector of the current conversation and calculates the stall probability through forward propagation. The server determines whether to trigger conversation topic generation by comparing the result with a threshold, thus replacing simple rule matching with probabilistic judgment at the technical level, reducing the false positive rate and improving detection accuracy.

[0182] In another implementation, the server uses a database management system to store user usage history. The server can maintain multiple or different types of tables for each user, such as a message table, a click behavior table, a favorites behavior table, and a recommendation feedback table. The message table may include fields such as user identifier, session identifier, message content, and sending time. The click behavior table may record fields such as recommended content identifier, click time, and dwell time. The server can periodically use a data processing library to statistically analyze each user's access frequency, feedback score, and recent activity level across different content categories in a background task. The server can then convert these statistical results into interest vectors, for example, assigning a weight to each tag and normalizing these weights before storing them in a user profile table.

[0183] Before invoking the generative AI model, the server can construct prompts based on the user profile's interest attributes and the current conversation topic. Instead of directly sending the current message to the model, the server combines information such as "user interest attributes," "historical viewing history summary," "current conversation background," and "generation task constraints" into a structured natural language description. The server can use predefined templates to populate different dimensions of information into corresponding slots. For example, the server can generate the following prompts: User's past viewing history: 1. Inception 2. Interstellar Travel 3. Blade Runner 2049 Current user message: 'Any good movies to watch lately?' Based on the above viewing history, please recommend 3 movies suitable for this user.

[0184] Require: 1. Provide the Chinese title and English title for each film (if applicable).

[0185] 2. Each film should be accompanied by 1-2 brief reasons for recommendation.

[0186] 3. The reasons for recommendation should include discussion points suitable for starting a conversation (such as interpretation of the ending, character relationships, value conflicts, etc.). Alternatively, the server can generate the following prompt: "Context of the conversation: Two friends are chatting, and one of the users says, 'I feel like we've run out of things to talk about.'"

[0187] This user's interest tags include: travel, food, and science fiction movies.

[0188] Please generate 5 questions or topics that can be used as a chat starter for this user, each topic not exceeding 30 characters.

[0189] Require: 1. Combine user interest tags.

[0190] 2. The topic should be natural and conversational, suitable for sending directly in chat applications.

[0191] 3. Try to make it easy for the other person to respond and share their experiences. The aforementioned prompts are automatically generated by the server using a specific algorithm. By filling template slots based on interest vectors and contextual features, the server enables the generative AI model to receive richer and more structured contextual input during the inference phase, thereby improving the match between the generated output and user needs. This data-driven prompt construction method differs from manually written prompts; instead, it relies on statistical analysis and feature extraction, making data flow and information expression within the computing system more efficient.

[0192] In a preferred implementation, the server employs a generative artificial intelligence model as its text generation engine. The server can deploy a multi-layered self-attention network based on a Transformer architecture on inference nodes equipped with graphics processing units. This network may include an embedding layer, several multi-head self-attention layers, a feedforward network layer, and a normalization layer. During the training phase, the server can pre-train and fine-tune the model using a large-scale dialogue corpus and recommendation-related corpora. The training process can employ masked language modeling, causal language modeling, or instruction-following tasks. The server can use adaptive optimization algorithms (such as Adam or its variants) to minimize the cross-entropy loss between the predicted output and the training objective, progressively updating the network weights. During the fine-tuning phase, the server can further introduce a dedicated dataset labeled with "conversation topics" and "recommendation reasons" to make the model more stable in topic generation and recommendation expression.

[0193] During inference, the server inputs the aforementioned prompts into the generative AI model, where they are encoded into a sequence of words by a word segmenter. The server can set parameters such as maximum output length, temperature parameters, and top-k or top-p sampling parameters to control the diversity and determinism of the generated content. After receiving the model output, the server uses a rule-based filtering module and a sensitive word recognition module to post-process the results, removing segments that do not conform to content specifications or exceed length limits, and performing repetition detection and sorting to output a structured list of conversation topics or a list of recommended information. In this way, the server internally forms a complete technical pipeline from the user's original input to highly relevant output. This pipeline achieves a stable improvement in the quality of the generated results through a combination of model architecture, parameter settings, and post-processing rules.

[0194] In terms of link metadata processing, the server can use a hypertext retrieval library to obtain the raw hypertext content from external resources linked in user messages. The server can use a hypertext parsing library to parse the document structure, extracting title text, summary text, and preview image addresses from title tags, description meta tags, and open graph meta tags. The server can perform language detection and length control on the extracted text, achieving unified encoding and truncation. The server can store this metadata in a link metadata table and associate link identifiers with dialog identifiers. When returning a response to the terminal, the server can encapsulate the title, summary, and image address corresponding to the link into preview data, enabling the terminal to render it as a link card for display to the user. Through this structured metadata processing, the server reduces the need for repeated requests to external resources on the terminal side, thereby reducing communication load and shortening content display latency on the user end.

[0195] In this embodiment, the terminal is primarily responsible for presenting the user interface and collecting user actions. The terminal can use a local UI framework to build the conversation interface, recommended topic panel, and link preview cards. After receiving the conversation topic array or recommended information array returned by the server, the terminal can group and display them according to type, for example, displaying "Movie Recommendations" and "General Chat Topics" in different areas. When a user clicks on a topic card, the terminal can automatically fill in the topic text into the input box, allowing the user to easily initiate a topic with a simple tap. The terminal can also send this click event as interest feedback to the server for subsequent updates to the user's interest attributes.

[0196] When using this system, users no longer need to manually search for topics or browse through large amounts of irrelevant information. Users simply type natural language messages on their terminals as usual, such as "Any good movies to watch lately?" or "I feel like I'm running out of things to talk about." The server can then automatically generate conversation topics and recommendations related to the user's interests based on their past usage history and current conversation state, presenting them in a format suitable for the terminal's display. This approach not only reduces the cognitive burden on users in topic selection but also improves the coherence of the conversation and the efficiency of information acquisition.

[0197] This invention's system goes beyond simply automating traditional manual recommendation processes; it improves upon traditional methods by refining the computer's internal data structure design, model invocation methods, prompt statement construction algorithms, and metadata processing workflows. The server encodes user interest attributes into vectors and explicitly injects these attributes into prompt statement templates, enabling generative AI models to utilize additional semantic signals during inference. This improves recommendation accuracy without increasing the burden on user interaction. Furthermore, the server uses probabilistic detection of session pauses to avoid frequent and meaningless calls to large model inference, reducing ineffective computations and lowering overall computational resource consumption, thus achieving a comprehensive optimization of computational efficiency and response time.

[0198] Furthermore, the server can employ data augmentation strategies during training and inference, such as paraphrasing and sentence transformation of user history messages to increase the diversity of training samples and thus improve the model's robustness to different expressions. During offline training, the server can use mini-batch training and learning rate annealing strategies to improve convergence speed and employ early stopping on the validation set to prevent overfitting. These combined techniques result in a more stable and reliable performance of the generative AI model within this system.

[0199] In other alternative implementations, the server can also employ different types of generative models, such as sequence-to-sequence structure-based models, hybrid models based on retrieval-enhanced generation, or introduce a multi-task learning architecture to simultaneously learn "topic generation," "recommendation reason generation," and "summary generation" tasks within the same model. The server can fuse the outputs of different modules through attention mechanisms or feature concatenation to further enhance the diversity and information density of the output content.

[0200] Through the aforementioned implementations, the server, terminal, and user achieve technical collaboration within the system: the server, through efficient data storage structures, feature extraction algorithms, prompt generation logic, and generative artificial intelligence model reasoning processes, achieves accurate modeling and response to session states and user interests; the terminal, through optimized display structures and interaction logic, presents complex background processing results in an intuitive manner; and users continuously provide the system with new data sources through simple natural language input and click operations. This system achieves a deep integration of natural language processing and recommendation technology in the field of computer technology, effectively improving the response quality, processing efficiency, and resource utilization of the information exchange system.

[0201] use Figure 12 The processing procedure is explained.

[0202] Step 1: Users type and send messages on the terminal.

[0203] Users input natural language text via the keyboard in the terminal's chat interface, such as "Any good movies lately?" or "I feel like we've run out of things to talk about," and then click the send button. The input consists of the text string entered by the user and the relevant action selected by the user (e.g., clicking a recommendation card). The terminal receives this input, encodes it into UTF-8 text data, and appends user identifier, session identifier, timestamp, and terminal type information to form a structured data object. The terminal then sends this structured data as a request message to the server based on network communication protocols. The output is a request data packet containing the aforementioned fields.

[0204] Step 2: The server receives and parses messages from the terminal.

[0205] The server receives request data packets sent by the terminal over the network, with the structured request data from step 1 as input. The server uses a network framework to parse HTTP or WebSocket messages, extracting fields such as user identifier, session identifier, message text, and timestamp, and then calls the database access module to write the message record to the message table. During the write operation, the server generates an internal message ID for the message and records the reception time. The server also records basic information about this request in the log system. The output consists of a message record stored in the database and a message object in memory for subsequent processing.

[0206] Step 3: The server constructs session characteristics based on the dialogue history and current messages.

[0207] The server reads the most recent message records of the current session from the database, with the input being the current message object and a list of related historical messages. The server calls the natural language processing module to segment and lexically represent the text of each message, and then uses a pre-trained semantic model to map it into a vector representation. The server also calculates features such as the time interval between messages, message length, and whether it contains questions. The server combines these textual and numerical features into a feature vector representing the current session state. The output is a session feature vector and contextual summary information used for subsequent decision-making.

[0208] Step 4: The server detects whether the session is stagnant or lacks topic leads.

[0209] The server takes the session feature vector generated in step 3 as input and feeds it into a pre-trained classification model (e.g., logistic regression or gradient boosting tree model). The server calculates the session stall probability and the lack of topic cues through forward computation. The server compares this probability to a threshold; if the probability is greater than or equal to a preset threshold, the server determines that the current session needs topic assistance. Based on the determination, the server generates a status label (e.g., "Topic generation needed" or "Topic generation not needed"). The output is the session status label and the stall probability value used in subsequent processes.

[0210] Step 5: The server extracts and updates the user's interest attributes.

[0211] The server takes the user's identifier as input and reads the user's historical behavior data from the database's usage history tables (including click tables, favorites tables, viewing records tables, etc.). The server uses a data processing library to statistically analyze these records, calculating the access frequency, recent activity, and feedback rating for different content categories (such as movies, travel, food, etc.). The server generates an interest weight vector based on a preset algorithm (such as a weighted average or time decay model) and extracts tags with higher weights as interest attributes. The server writes the new interest attributes back to the user profile table and retains the list of interest tags required for the current session in memory. The output is the updated set of user interest attributes and their corresponding weights.

[0212] Step 6: The server provides prompts for constructing generative artificial intelligence models.

[0213] The server takes the current message text, a summary of the conversation history, and a set of interest attributes as input, and selects an appropriate prompt template (such as a movie recommendation template or a chat topic template). The server inserts information such as the user's past viewing history, interest tags, current message content, and generation goal (generating a conversation topic or recommendation information) into the reserved slots in the template. The server also determines whether to add a "conversation paused" or "opening topic needed" message to the prompt statement based on the conversation status flags. Through this string concatenation and placeholder replacement operation, the server generates a complete prompt statement text. The output is a prompt statement containing context, interest information, and generation requirements.

[0214] Step 7: The server calls a generative artificial intelligence model to generate data for information exchange.

[0215] The server takes the prompt generated in step 6 as input and passes it to the generative AI model service deployed on the graphics processing unit. The server first uses a tokenizer to convert the prompt into a sequence of words, and then the model performs multi-layer self-attention computation to generate the probability distribution of the output sequence. The server progressively generates text output based on the set maximum length, temperature parameters, and sampling strategy (such as top-k or top-p). The server receives the complete text result returned by the model and splits it into multiple conversation topics or recommendation items using parsing rules. Each item contains fields such as title, description, or recommendation reason. The output is a structured list of conversation topics or recommendation information.

[0216] Step 8: The server filters and reshapes the results to generate candidate information.

[0217] The server takes the list of conversation topics or recommendations obtained in step 7 as input and performs content filtering and formatting on each entry. The server uses a sensitive word list and rule engine to detect and remove inappropriate content; it detects and merges duplicate or highly similar entries; and it sorts the entries according to user interest weights, placing entries with high match rates to high-weight interest tags at the top. The server also truncates the text length according to terminal display constraints and organizes the fields to form a candidate information data structure for client display. The output is a list of filtered and sorted candidate conversation topics or recommendations.

[0218] Step 9: The server parses the link information in the user message and generates metadata.

[0219] The server takes the original user message text as input and identifies link information using regular expressions or URL detection modules. For each detected link, the server uses an HTTP client library to send a request to the corresponding external resource to retrieve the raw data returned by the webpage or API. The server uses an HTML parsing library to extract title tags, description meta tags, and image meta tags from the webpage, or to extract title fields, summary fields, and image fields from the structured data returned by the API. The server performs length control and encoding uniform processing on the extracted text and encapsulates the obtained title, summary, and image information into metadata objects. The output is a metadata collection corresponding to each link.

[0220] Step 10: The server stores data and metadata used for information exchange and generates response messages.

[0221] The server takes the candidate information generated in step 8 and the metadata generated in step 9 as input. It associates this data with the current session ID and user ID, storing it in the generated result table and metadata table of the database for subsequent querying and analysis. The server then constructs a response message, combining the candidate session topics, recommendation information, and link preview metadata into a unified response structure. The server includes necessary identification fields (such as item ID and type markers) in the response for the terminal to use in subsequent feedback. The output is structured response data tailored to the terminal.

[0222] Step 11: The terminal receives the server's response and presents candidate information and a link preview.

[0223] The terminal takes the response data from step 10 as input, receives the message returned by the server through the network communication module, and parses the candidate topics, recommendation information, and metadata fields. The terminal invokes local user interface components to render candidate conversation topics as topic cards or recommendation bubbles, displays recommendation information in a list format, and renders link metadata as preview cards with titles, summaries, and thumbnails. The terminal adjusts the layout and fonts according to screen size and interface style, allowing users to intuitively browse and select this information. The output is an updated interface display, including multiple clickable candidate topic cards and link preview elements.

[0224] Step 12: Users select and send new session content based on candidate information.

[0225] Users input candidate topic cards or recommended items displayed on the terminal interface, triggering the terminal's interaction logic by clicking on a card or item. Upon receiving the click event, the terminal automatically fills the corresponding topic text into the input box, which the user can send directly or after minor modifications. The terminal then sends this new message back to the server as new input, and can simultaneously upload information about which candidate item was clicked as feedback. The output consists of new user messages and interest feedback data, forming a closed loop within the system, enabling the server to continuously optimize interest attributes and generation strategies in subsequent steps.

[0226] 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.

[0227] 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."

[0228] Existing dialogue assistance technologies based on generative AI models typically generate responses directly from a limited context within the current conversation, lacking analysis and utilization of users' long-term behavioral patterns. As a result, on the one hand, servers cannot accurately determine when a user is truly "speechless" or "unable to initiate or continue information exchange"; on the other hand, the prompts provided by the server to the generative AI model are often crude, lacking structured constraints tailored to specific user behavioral characteristics and interests, leading to mismatches between generated results and user interests, low topic relevance, or inconsistent style.

[0229] Furthermore, in traditional systems, servers typically treat generative AI models as a "black box," simply forwarding inputs and outputs without considering user history, active time periods, or terminal interface status for comprehensive judgment. They also fail to systematically filter or structure the generated results. This not only leads to low server resource utilization (e.g., frequently calling the generative model when users don't actually need topic suggestions) but may also generate redundant or inappropriate topic suggestions, thus failing to effectively improve the user's actual online communication experience.

[0230] Furthermore, due to the lack of a precise detection mechanism for users' "lack of information exchange cues" and a corresponding procedural processing flow, existing systems struggle to achieve an integrated, automated control chain at the computer level, encompassing "behavioral pattern modeling → state detection → instruction generation and construction → invoking generative AI models → result filtering and push." ​​Therefore, how to introduce user behavior pattern analysis methods on the server side, and on this basis, automatically construct high-quality prompts to drive generative AI models to generate suitable dialogue topics on demand, thereby improving the efficiency and stability of online information exchange at the computer system architecture and information processing flow levels, has become a pressing technical challenge.

[0231] 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.

[0232] In this invention, the server includes means for exchanging information between user terminals via a communication processing device; means for obtaining the user's information sending history from a storage device based on user identification information; means for parsing the sending time and content-related data contained in the information sending history on a computing processing device to extract the user's behavioral patterns in information exchange; means for determining, based on the behavioral patterns, current time, and terminal interface status information, whether the user has not sent information within a predetermined time period and has not sent information while the information input interface is open, thereby detecting that the user lacks information exchange clues; means for generating a prompt statement based on the detection result and the behavioral pattern to instruct a generative artificial intelligence model to generate topic data presumed to be of interest to the user; means for inputting the generated prompt statement into the generative artificial intelligence model and causing the generative artificial intelligence model to generate topic data as information exchange clues; means for filtering and formatting the generated topic data according to a predetermined standard and sending it to the user terminal as notification data; and means for recording the information sending history and the topic data in the storage device. This allows for an integrated computing process within the server, encompassing user behavior data collection, behavior pattern modeling, detection of missing topics, construction of structured prompts, generation of topic data by a generative AI model, and filtering and delivery of the generated results to the user terminal. From a computer technology perspective, this precisely invokes the generative AI model on demand when the user truly needs dialogue clues, improving the quality and relevance of prompt input, enhancing the consistency between generated topics and user interests and behavior patterns, thereby significantly improving the continuity and efficiency of online information exchange, and optimizing resource utilization and user interaction experience of the entire information processing system.

[0233] "Communication processing device" refers to an electronic device or its functional module used for data transmission and reception, session establishment and maintenance between multiple user terminals and servers, and can forward messages, control connection status and ensure the reliability of information transmission through wired or wireless network protocols.

[0234] "User terminal" refers to an electronic device operated by a user for information input, display and interaction, including but not limited to mobile terminals, fixed terminals, laptop devices or other devices with communication and interface display functions.

[0235] "Storage device" means a computer-accessible storage medium used to store program, user data, log data and model data in a read-write manner, including but not limited to semiconductor storage media, magnetic storage media, optical storage media or combinations thereof.

[0236] "Message sending history" refers to a set of sending behavior data recorded by the server within a predetermined time range and associated with a specific user. It includes at least one sending timestamp, message content information, and dialogue object information.

[0237] "Computational processing unit" refers to hardware resources and their control units that execute program instructions to perform data parsing, feature extraction, model reasoning, and logical judgment, including but not limited to central processing units, graphics processing units, dedicated accelerators, or combinations thereof.

[0238] "Behavioral patterns" refer to a set of data structures or parameters extracted from the history of information sending that reflect the regular characteristics of users' time distribution, frequency, content and topic preferences during the information exchange process.

[0239] "Terminal interface status information" refers to data reported by the user terminal or maintained by the server that indicates the current display and interaction status of the terminal interface, including but not limited to whether the chat window is open, whether the input box is active, the cursor activity status, and the most recent interaction time.

[0240] "Lack of information exchange cues" refers to a state determined based on behavioral patterns and the current user operation status, in which the user is in a potential communication scenario but does not generate new sending behavior within a predetermined time, and it is difficult to naturally start or continue the conversation.

[0241] "Generative AI models" refer to AI models that can automatically generate natural language text or other forms of content based on input text or other structured data. They are usually based on deep learning structures and learn parameters through large-scale training data.

[0242] "Prompt statements" refer to natural language or natural language-like instruction texts constructed by the server and input into the generative artificial intelligence model, used to instruct the generative artificial intelligence model to generate specific types or styles of output content under given constraints.

[0243] "Topic data" refers to a collection of text information generated by generative artificial intelligence models based on prompts, which can serve as a starting point or continuation clue for information exchange between users. It usually manifests as one or more topic sentences or topic suggestions that can be directly used in conversation.

[0244] "Filtering and formatting" refers to the post-processing operations performed by the server on the topic data output by the generative artificial intelligence model, including but not limited to length limits, sensitive content removal, duplicate content removal, sorting, and reorganization according to a predetermined interface layout.

[0245] "Notification data" refers to message data sent by the server to the user terminal to notify the user of topic suggestions or other auxiliary information. It includes topic content to be displayed on the terminal interface and related identification information or metadata.

[0246] In one embodiment of the present invention, the server is set up in a data center or cloud computing environment, and includes a computing device, a storage device, and a communication processing device. The operating system running on the server can be a general-purpose server operating system, such as a server operating system based on a Unix-like kernel. Backend applications are deployed on the server, and these applications can be implemented using a general-purpose programming language, such as an application service environment composed of an interpreted language and a corresponding network framework.

[0247] The server stores program instructions and data structures corresponding to multiple functional modules in its storage device, including at least: a session management module, an information sending history management module, a behavior pattern parsing module, a generative artificial intelligence invocation module, a prompt statement generation module, and a result filtering and notification generation module. The server establishes a secure connection with the terminal through a communication processing device, which can be a smart mobile terminal, a personal computing device, or other device with a human-computer interaction interface. Users access the services provided by the server through the user interface application on the terminal.

[0248] In communication between the server and the terminal, the server utilizes reverse proxy software and secure transmission protocols to receive and forward requests. The terminal sends user input, interface status events, and session control commands to the server via secure transmission protocols or persistent connection protocols. The server maintains information such as session identifiers, authentication tokens, and the most recent interaction time associated with each terminal in its session management module.

[0249] In the message sending history management module, when a user sends chat messages or other exchanged information through a terminal, the server saves the received messages in a structured record format to a relational database. Each record includes at least: user identifier, dialogue identifier, sending timestamp, message content text, message length, recipient identifier, and possibly topic tag fields. The server indexes these records in chronological order and maintains a cache of the most recent records for each user to facilitate fast access by the behavior pattern analysis module.

[0250] In the behavior pattern parsing module, the server uses a machine learning framework to load a pre-trained behavior analysis model. This model can employ a deep neural network structure, such as a multilayer perceptron or a one-dimensional convolutional neural network, to model time-series features; or it can use a recurrent neural network structure (such as a Long Short-Term Memory network LSTM) or gated recurrent units to model the time dependence of user sending behavior. The server converts the raw data from the message sending history into feature tensors, including: histograms of sending frequency divided by time intervals, sending time series over several hours or days, message length statistics, and topic tag distribution. The server normalizes and encodes these features (e.g., mapping time to periodic features) and inputs them into the behavior analysis model.

[0251] Within the behavior analysis model, the server performs multi-layer feature extraction on the input features through operations such as multi-level linear transformations, non-linear activation functions, and batch normalization. When training this behavior analysis model, the server uses supervised or semi-supervised learning methods: during the training phase, the server labels each time period with tags such as whether it belongs to a user's "regularly active period," "high-probability sending period," or "low-probability sending period," using cross-entropy loss or mean squared error loss as the objective function, and iteratively updates the network weights through gradient descent and its variants (such as adaptive learning rate algorithms). When necessary, the server uses data augmentation methods, such as perturbing the timeline and adding noise to message statistical features, to improve the model's robustness to fluctuations in behavior patterns.

[0252] When the server performs inference on the behavioral analysis model at runtime, it only performs forward propagation operations, generating a set of parameters to represent user behavior patterns. For example, the server obtains a vector of typical active time periods for users, a statistical distribution (mean, standard deviation) of users' usual message intervals, and a probability distribution of user-preferred topic categories. The server combines these outputs into a behavioral pattern data structure, stores it in memory, and makes it accessible to subsequent modules.

[0253] In terminal interface state management, the server receives interface events reported by the terminal, such as "chat window open," "input box active," "user typing," and "window minimized." When a user opens a session window, the terminal sends an event message to the server, which records this event along with a timestamp as terminal interface state information. The server can use memory storage or a caching system to maintain each user's current interface state and its duration.

[0254] In detecting a lack of information exchange cues, the server reads the behavioral pattern data structure and the current time, and combines this with the terminal interface status information to make logical judgments. The server calculates the interval Δt between the current time and the time of the user's most recent message, and compares Δt with the "normal interval threshold" stored in the behavioral pattern. For example, the server can use a rule: when Δt is greater than a certain multiple of the normal interval (e.g., 3 times), and the current time output by the behavioral analysis model belongs to a "high-probability sending period," and the terminal interface status shows that the user has opened the input interface and has not generated a new message record for more than a preset time window (e.g., 60 seconds), the server marks this state as "lacking information exchange cues." This detection is not simply based on a fixed time threshold, but combines individualized behavioral pattern statistics, enabling the server to more accurately identify the moments when the user truly needs topic assistance, thereby reducing unnecessary model calls and lowering communication and computational load.

[0255] In the prompt generation module, after detecting the above situation, the server extracts information such as the user's interest category (e.g., technology, movies, travel), recent conversation topic summary, and conversation partner type (e.g., friends, colleagues) from the behavior pattern data structure. The server then uses a predefined text template to insert this structured data into template placeholders, forming a prompt in natural language. The server can generate different types of prompts based on the application scenario. For example: "Based on the following user information and chat background, please generate 3 suggestions for this user that can naturally initiate or continue the conversation:" User interests: Technology, Movies, Travel Recent chat topics: Movies watched last week, vacation plans for next month. Require: 1. Each suggestion should not exceed 50 words; 2. The tone is friendly and natural; 3. Do not involve sensitive or controversial topics.

[0256] Please directly input a topic sentence in Chinese that you can send to the recipient. or: "If a user stays in the chat window for a long time but does not send a message, please generate 5 chat topic openings based on recent news for that user."

[0257] Require: 1. The topic should be related to technology or daily life; 2. Suitable for easy chatting among friends; 3. Each opening statement should be limited to one sentence.

[0258] Please output in Simplified Chinese. as well as: Based on the following summary of past conversations, please provide the user with three suggestions for continuing the conversation: Historical Summary: The two sides had just finished discussing their respective work situations, and the atmosphere was relaxed.

[0259] Require: 1. Offer follow-up questions or new topics that the other party might be interested in; 2. Try to add a touch of humor, but don't be offensive; 3. Each suggestion can be directly copied and sent to the other party.

[0260] Please output the result in Chinese. When constructing prompts, the server includes not only content direction constraints (interest categories, historical topics) but also parameters such as the number of prompts, length limits, tone requirements, and prohibited topics. This structured prompt structure subjects the generative AI model to stricter constraints during internal decoding, thereby reducing the generation of irrelevant content and improving the relevance and usability of the generated results. This procedural generation of prompts is essentially a technical constraint and optimization of the generative model's input space.

[0261] In the generative AI invocation module, the server encapsulates the prompt statement as input text, along with model configuration parameters (such as temperature, maximum generation length, and number of candidates), into request data, and sends it to the inference service deployed with the generative AI model via a communication processing device. The generative AI model can employ a large-scale language model based on the Transformer architecture, internally including multi-layered self-attention networks, feedforward networks, and embedding layers. When training this model, the server can pre-train it using a large-scale text corpus and fine-tune it in dialogue or topic suggestion scenarios, using a cross-entropy loss function combined with masked language modeling or autoregressive generation strategies. During the inference phase, to save computational resources, the server can employ sampling strategies commonly used in dialogue scenarios (such as top-k or top-p sampling) and decoding pruning methods to reduce verbose and repetitive output.

[0262] After receiving topic data from the generative AI model, the server performs a series of post-processing operations in the result filtering and notification generation module. The server checks the length of each generated text, truncating or discarding text exceeding a predetermined word count; it identifies and removes inappropriate content using a sensitive word filter list or content compliance detection module; and it scores and ranks candidate topics based on their match with user interests, prioritizing topics with higher relevance. Finally, the server packages the retained topics into a notification data structure, including the topic text, suggestion type identifier, generation timestamp, and other metadata.

[0263] The server sends notification data to the corresponding terminal via a communication processing device. Upon receiving the notification data, the terminal renders a list of suggested topics or speech bubbles in its user interface. The terminal can provide click functionality; when a user clicks on a candidate topic, the terminal automatically fills in the topic text into the input box, allowing the user to send or edit it directly. The terminal can visually distinguish between "system-recommended topics" and user-inputted content to improve interface readability.

[0264] Users input information on their terminals, referring to topic suggestions provided by the server, and then resend the message through the terminal. The server adds the new message to the message sending history, and the behavior pattern analysis module uses the updated historical data in subsequent analyses to gradually form a dynamic model of user behavior. Through this closed loop, the server continuously adjusts the behavior pattern parameters and the threshold for detecting the lack of topic states, enabling the system to adapt to changes in user behavior.

[0265] Through the aforementioned structure and processing flow, the server does not simply automate the execution of human-generated topics. Instead, it achieves a technical solution optimized for both computing and communication resources by statistically modeling the time-series characteristics of user behavior, comprehensively judging the terminal interface state, structurally constructing prompts, constraining the decoding space of the generative artificial intelligence model, and filtering and sorting the generated results. Since the server only invokes the generative artificial intelligence model when it detects statistically significant "abnormal pauses" and the interface state indicates that the user is in an input scenario, it can significantly reduce invalid calls, lower average response latency, and reduce network load. Simultaneously, through the construction of personalized prompts driven by behavior patterns, the server enables the output of the generated model to have higher consistency in terms of topic relevance and user acceptance, reducing the workload of users manually filtering irrelevant topics and improving overall system-level information processing efficiency.

[0266] In alternative implementations, the server can employ different behavioral analysis model structures. For example, it can model the relationships between users, dialogue partners, and topics based on graph neural networks to extract higher-level social relationship features; or it can use a hybrid model that combines rule-based anomaly detection (such as fixed time thresholds) with neural network-based pattern recognition to further improve detection accuracy. The server can also introduce multilingual support fields in the construction of prompt statements, instructing the generative artificial intelligence model to generate topics in different languages, thereby enabling applications in cross-language scenarios.

[0267] Through the aforementioned hardware and software components, along with a specific data processing flow based on generative artificial intelligence models and prompt statements, the server achieves technical optimization of the user's online information exchange process. This not only improves the user's subjective experience but also integrates behavioral pattern modeling, on-demand resource allocation, and post-processing of results within the computer, resulting in verifiable technical effects in terms of processing speed, topic relevance, data management, and communication load.

[0268] use Figure 13 The processing procedure is explained.

[0269] Step 1: The server establishes a secure connection with the terminal through a communication processing device and receives user identification information.

[0270] Input: Login request sent by the terminal, user identification information, session establishment request.

[0271] Output: Session ID, authentication token, connection channel information.

[0272] After receiving a login or chat interface request from a terminal, the server parses the user identification information (such as user ID or token) in the request message, allocates or updates a session identifier in the session management module, and registers the session's state in memory. The server encapsulates the generated session identifier and authentication result into response data and returns it to the terminal via a secure transmission protocol for subsequent message and status reporting.

[0273] Step 2: The terminal generates a message sending request under user operation and submits the message content to the server.

[0274] Input: The text content entered by the user in the terminal interface, the current session identifier, and the target session object identifier.

[0275] Output: Send message request message.

[0276] After a user types a message in the chat input box on the terminal and clicks send, the terminal encapsulates the message text along with the current session identifier, sending timestamp, and dialogue partner identifier into a request, and sends it to the server using a secure protocol or persistent connection protocol. The terminal records the message's sending status locally (pending confirmation, sent, etc.) to facilitate retrying in case of network anomalies.

[0277] Step 3: The server receives and stores the user's information sending history.

[0278] Input: The message request message sent by the terminal (including message content, timestamp, user identifier, dialogue object identifier, etc.).

[0279] Output: The information transmission history written to the storage device, and the updated last transmission time cache.

[0280] Upon receiving a message request, the server parses the message fields and writes the user identifier, conversation identifier, sending timestamp, message length, and message text into the message record table of a relational database. Simultaneously, the server updates the user's "last sent time" and "last conversation partner" in memory or a cache system for subsequent behavioral pattern analysis and status detection.

[0281] Step 4: The server periodically reads each user's historical sending records from the storage device.

[0282] Input: Scheduled task trigger signal, list of user IDs, query time range (e.g., the last 7 days).

[0283] Output: Send historical data sets (list or time series structure) according to the information organized by the user.

[0284] When the scheduled task arrives, the server performs a database query for each online or monitored user. Based on the user identifier and time range, it retrieves all sending records for that user within the specified time period. The server sorts the records by sending time, constructs the results into a list or time series object containing fields (timestamp, message length, topic tags, etc.), and stores it in an in-memory data structure to provide input for the behavior pattern analysis module.

[0285] Step 5: The server extracts features from the information sending history and constructs them into the input tensor of the behavior analysis model.

[0286] Input: The set of historical data on user information sent in step 4.

[0287] Output: Behavioral analysis feature tensor or feature vector set.

[0288] The server iterates through the historical record list, calculating statistical features at different time granularities, such as the number of messages sent per hour, daily active time periods, average message length, and the distribution of time intervals between adjacent messages. The server maps timestamps to numerical features (e.g., normalizing the time of day to [0,1] or splitting it into "hour" and "weekday / non-weekday" features), and stores these features as fixed-length vectors or matrices according to a predetermined window size. The server then packages the resulting feature set into a tensor data structure, which serves as input to the behavioral analysis model.

[0289] Step 6: The server uses behavioral analysis models to infer and calculate user behavior patterns.

[0290] Input: Behavioral analysis feature tensor generated in step 5, and pre-loaded behavioral analysis model parameters.

[0291] Output: Behavioral pattern data structure (including active time period parameters, normal interval thresholds, topic preference distribution, etc.).

[0292] The server, on its processing unit, invokes the inference interface of a machine learning framework, inputting feature tensors into the behavioral analysis model. Internally, the model uses multi-layer linear transformations, adaptive activation functions, and possible recurrent or convolutional structures to calculate the user's activity probability at different time periods, typical message interval statistics, and preference probabilities for each topic category. The server organizes the multiple numerical vectors output by the model into structured data, such as a list of active time periods, the average and standard deviation of typical intervals, and the probability distribution of topic categories, and saves them as a behavioral pattern data structure in key-value pair format.

[0293] Step 7: The terminal reports the current interface state to the server to reflect the user's interaction context.

[0294] Input: UI events triggered by the terminal (chat window opens, input box is activated, user is typing, etc.), current timestamp.

[0295] Output: Interface status reporting message.

[0296] When a user opens a chat window or clicks an input box on the terminal, the terminal captures these interface events, along with the user identifier, session identifier, and event timestamp, and sends them to the server via a persistent connection or request-response method. The terminal then updates the current interface state locally for interface rendering and interaction logic.

[0297] Step 8: The server maintains and updates the terminal interface status information for each user.

[0298] Input: The interface status reporting message sent by the terminal in step 7.

[0299] Output: Interface state records stored in memory (including the most recent state and its duration).

[0300] After receiving the interface status report, the server updates the corresponding user's status field in the in-memory status table based on the user identifier and session identifier. For example, it marks "Current window is open", "Input box is active", and "Typening" and records the status effective time. The server dynamically maintains the duration of the interface status by calculating the difference between the current time and the status effective time, which is used to determine the situation where there is a lack of information exchange clues.

[0301] Step 9: The server detects whether a user is in a state of lacking information exchange cues based on behavioral patterns and interface status.

[0302] Input: Behavioral pattern data structure obtained in step 6, interface state record maintained in step 8, and user's most recent sending time.

[0303] Output: Lack of information exchange clues (Boolean value), score for possible abnormal pauses.

[0304] The server calculates the time interval Δt between the current time and the user's most recent message time, and compares it with typical interval statistics in the behavior pattern (such as the mean μ and standard deviation σ). For example, it determines whether Δt is greater than μ + K·σ (where K is a preset coefficient). If Δt is significantly larger and the behavior pattern indicates that the current time belongs to a high-activity probability interval, and the interface status shows that the user has opened a chat window and kept the input box active for more than a set threshold T1, then the server marks this status as "lacking information exchange clues". The server can further calculate a score based on the degree of deviation of Δt from the statistical threshold, representing the degree of pause abnormality, which is used to adjust the number or type of subsequent generated topics.

[0305] Step 10: When the server detects a lack of information exchange clues, it extracts the context information needed to generate a prompt statement.

[0306] Input: Lack of information exchange clues flags from step 9, behavioral pattern data structure, and recent dialogue history summary data.

[0307] Output: The set of context parameters required to construct the prompt statement (interest category, recent topics, dialogue object type, generation constraints).

[0308] When the flag is true, the server reads the user's list of interest categories from the behavioral pattern (e.g., selecting several categories with higher probabilities based on the probability distribution of topic categories) and extracts topic summaries from several recent conversations, which can be obtained through simple rules (such as keyword statistics) or a pre-trained text summarization module. The server also determines the type of the conversation partners (e.g., based on the relationship tags between the conversation partners) and determines tone style and sensitive topic restriction rules. The server combines this information into prompt statement construction parameters, such as: interest categories, recent topic summaries, expected number of prompts N, maximum length of a single prompt L, tone description, and set of prohibited topics.

[0309] Step 11: The server constructs prompts for generative artificial intelligence models based on context parameters.

[0310] Input: The set of context parameters obtained in step 10, and a predefined text template.

[0311] Output: The complete prompt text.

[0312] The server selects a prompt template that matches the current scenario, such as a template suitable for "continuing the conversation" or a template based on a news opening. The server inserts context parameters into placeholders in the template and generates a prompt in natural language format, such as the aforementioned: "Based on the following user information and chat background, please generate 3 suggestions for this user that can naturally initiate or continue the conversation:" User interests: Technology, Movies, Travel Recent chat topics: Movies watched last week, vacation plans for next month. Require: 1. Each suggestion should not exceed 50 words; 2. The tone is friendly and natural; 3. Do not involve sensitive or controversial topics.

[0313] Please directly input a topic sentence in Chinese that you can send to the recipient. During data processing, the server concatenates parameter values ​​with fixed text and performs basic format checks (such as whether the text is empty or whether the length exceeds the limit). Finally, it generates a continuous text string as a prompt statement output.

[0314] Step 12: The server calls a generative artificial intelligence model to generate topic data.

[0315] Input: The prompt text constructed in step 11, model call parameters (temperature, maximum length, number of candidates, etc.).

[0316] Output: Multiple topic text candidate results returned by the generative artificial intelligence model.

[0317] The server encapsulates the prompt and model parameters into a request message, which is then sent to the inference server endpoint, where the generative AI model is deployed, via a communication processing device. Internally, the inference service uses a multi-layered Transformer network to encode the prompt and generates several candidate text sequences word-by-word or sub-word-by-sub-word based on an autoregressive decoding strategy. Upon receiving the response, the server parses the returned JSON or other format, extracts each generated topic text, and forms the original topic candidate list.

[0318] Step 13: The server filters and formats the generated topic data.

[0319] Input: The original list of topic candidates obtained in step 12, predefined filtering rules, and formatting rules.

[0320] Output: A final list of topic data after filtering and formatting.

[0321] The server iterates through each candidate topic text sequentially. First, it checks if the text length is within a specified range, truncating or removing text exceeding the length limit. Then, it uses a sensitive word database matching or content moderation module to detect non-compliant words and topics, deleting candidates containing sensitive content. The server scores the topic based on its relevance to interest categories, for example, by calculating keyword overlap or using a simple classifier to determine the topic category. It then sorts the topics by score and selects the top-scoring ones as the final output. The server can also add metadata to each topic, such as the creation time and corresponding interest tags, for display on the terminal interface.

[0322] Step 14: The server encapsulates the final topic data into notification data and sends it to the terminal.

[0323] Input: The final list of topic data obtained in step 13, user ID, and session ID.

[0324] Output: Notification data message sent to the terminal.

[0325] The server organizes the topic text and its metadata into a notification structure, such as an array of topics, a text field for each topic, and a type field. The server sends this notification to the target terminal via a persistent connection or push service. The server logs this notification action in its internal logs for subsequent system performance monitoring and effect statistics.

[0326] Step 15: The terminal displays topic suggestions and provides interactive operations in the user interface.

[0327] Input: The notification data message sent by the server in step 14.

[0328] Output: Topic suggestion elements displayed on the terminal interface, and user-selectable click events.

[0329] After receiving the notification data, the terminal parses the candidate topic list and displays it in the chat interface as a list or speech bubble. The terminal adds an "Insert into Input Box" button to each topic; when a user clicks on a topic, the terminal automatically fills the text into the input box of the current conversation, allowing the user to send it directly or edit it further. The terminal can also display these topics as placeholder prompts above the input box to help users quickly select them.

[0330] Step 16: Users edit and send new messages based on recommended topics, and the server updates the behavioral data.

[0331] Input: The final message text, session identifier, and user identifier in the terminal input box.

[0332] Output: New message sending records and updated behavior pattern training data.

[0333] After viewing recommended topics, users can directly send recommended text, or modify or add their own content. The terminal sends the edited message to the server, which stores the message and adds it to the message sending history, similar to step 3. During the next behavior pattern analysis, the server incorporates this new data into feature calculations, allowing the behavior analysis model to gradually learn and adapt to the user's actual behavior under topic-assisted conditions. This further improves the accuracy and personalization of subsequent detection and prompt generation.

[0334] 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".

[0335] In existing information exchange and content recommendation technologies, although servers can push information to terminals based on static rules or simple recommendation algorithms, the following technical problems still exist: First, servers typically lack the ability to comprehensively model user behavior history, state information, and emotional information, resulting in the inability to trigger information generation processing in a timely and accurate manner at critical moments when users lack information clues or the starting point of the dialogue, thus interrupting the interaction process and reducing the utilization rate of system resources; Second, when calling generative artificial intelligence models, servers mostly rely on manually written fixed prompt text, and the prompt statements cannot be adaptively generated based on real-time behavioral data and emotional data, resulting in insufficient relevance of the generated results to the user's current intent and context, and the computational power of generative artificial intelligence models is not fully utilized; Third, the management of generated results by servers is mostly limited to one-time responses, lacking a mechanism for structured association and continuous feedback between generated data and user behavior history and emotional characteristics, and failing to form a closed-loop optimization process based on interest vectors, preference vectors, and emotional features, thereby limiting the improvement of the model in terms of recommendation quality, dialogue coherence, and overall system performance.

[0336] The technical challenge to be addressed by this invention is to provide a system structure that can automatically determine on the server side whether a user is in a state of lacking information clues or a starting point for dialogue, and dynamically generate prompt statements based on behavioral history information, attribute information, and emotional information to invoke a generative artificial intelligence model, and can provide bidirectional feedback between the generated data and behavioral history. This forms an end-to-end technical link within the computer for data acquisition, semantic parsing, prompt statement generation, model invocation, result selection, and feedback updates, thereby improving the server's processing efficiency, response accuracy, and resource utilization in information exchange, topic generation, and content recommendation scenarios, and improving the functionality and performance of the computer technology itself.

[0337] 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.

[0338] In this invention, the server includes a data acquisition and parsing unit for acquiring input information and parsing it into behavioral history information, attribute information, and emotional information; a state recognition unit for automatically determining whether the user is in a state without information clues or a dialogue starting point based on the behavioral information or state information; a prompt statement generation unit for extracting features from the information set containing behavioral history information, attribute information, and emotional information and generating prompt statements for input to a generative artificial intelligence model; a model invocation unit for inputting the prompt statements into the generative artificial intelligence model and obtaining generated data including information for information exchange, dialogue starting topic, or content recommendation information; and a model invocation unit for processing the generated data. The server includes a result management unit that associates and stores generated data with corresponding behavioral history, attribute, and emotional information; a feedback update unit that summarizes and preprocesses behavioral history information at predetermined time intervals, calculates interest or preference vectors and embeds them into prompt statements, and then evaluates, selects, and sorts the generated data; and a communication control unit that uses natural language processing methods to determine emotional information from text input, extracts candidate users based on behavioral history and emotional information, generates prompt statements that indicate the generation of messages or topics that promote information exchange or product proposals, and sends the message or topic information output by the generative artificial intelligence model to the corresponding terminal device. This creates a closed-loop processing flow within the server, from multi-source behavioral and emotional data acquisition to adaptive prompt statement construction, generative artificial intelligence model invocation, and result feedback optimization. This automates and finely schedules the information exchange control logic and recommendation logic, improves the server's parallel processing capabilities and response quality in multi-user, multi-scenario applications, and substantially improves the technical performance of computer systems in dialogue support, content recommendation, and user matching applications.

[0339] "System" refers to a computer technology whole consisting of at least one server, at least one terminal device, and communication and storage devices for realizing information exchange and data processing, wherein the constituent elements work together through a predetermined communication protocol to execute the method of the present invention.

[0340] A "server" refers to a computer device that includes processing, storage, and communication devices, and is a central node used to perform program processing such as data acquisition, parsing, prompt generation, generative artificial intelligence model invocation, and data management and feedback updates.

[0341] "Terminal device" refers to an electronic device operated by a user and communicating with a server, including but not limited to mobile terminals, fixed terminals or wearable terminals, used to send input information to the server and display generated data returned by the server to the user.

[0342] "Processing unit" refers to the hardware component in a server that executes computer programs, processes input information, and performs logical operations of various functional units, including but not limited to a central processing unit or a graphics processing unit.

[0343] "Input information" refers to various data sent by the terminal device and received by the server, including at least user behavior information, status information, text information, attribute information, and emotion-related information derived therefrom.

[0344] "Behavioral information" refers to data used to represent a user's operational behavior in the system, including but not limited to click records, browsing records, dwell time, playback control operations, and interface dwell state.

[0345] "Status information" refers to data used to indicate the user's current interaction state or the internal session state of the system, including but not limited to whether the user is in a chat interface, whether text has been entered, the duration of the interaction, and the session context identifier.

[0346] "Behavioral history information" refers to a collection of historical data obtained by recording and accumulating user behavior information within a predetermined time range, which is used to reflect the user's past operating patterns and usage habits.

[0347] "Attribute information" refers to static or semi-static characteristic data related to users, including but not limited to age range, interest categories, occupation categories, geographical regions, and preference tags.

[0348] "Emotional information" refers to the labeled information used to represent the user's psychological state, including emotion type and emotion intensity, obtained by performing sentiment analysis on the text or other data in the input information.

[0349] "Information clues" refer to the starting content or reference information that can guide users to continue to obtain information or exchange information, such as the next content to browse, related topics, or information on available options.

[0350] "Conversation start point" refers to a text message, topic, or question used to initiate or restart communication in a conversational context, prompting the user to send the first or subsequent message to the other party.

[0351] "Generative artificial intelligence models" refer to natural language generation models based on machine learning algorithms, especially deep learning, that can automatically generate semantically and contextually appropriate text data or recommendation information based on input prompts.

[0352] "Prompt statements" refer to text instructions that are automatically constructed by the server based on behavioral history information, attribute information, and emotional information and input into the generative artificial intelligence model, used to instruct the generative artificial intelligence model to perform specific information generation tasks.

[0353] "Generated data" refers to the output data of a generative artificial intelligence model after receiving a prompt statement, including but not limited to information text used for information exchange, dialogue starting topics, and content recommendation information.

[0354] "Content recommendation information" refers to the recommendation results and related explanatory information generated based on the user's interests, preferences, and behavioral patterns, for the objects that should be browsed, viewed, or consumed next.

[0355] "Storage device" refers to hardware resources used to persistently store behavioral history information, attribute information, emotional information, and generated data, including local storage devices or network storage devices.

[0356] "Communication device" refers to the hardware and its supporting software used to transmit data between a server and a terminal device, including network interface modules and control programs that implement communication protocols.

[0357] "Response information" refers to the feedback behavior data of users on the terminal device in response to the generated data, including but not limited to operation records such as clicking, ignoring, modifying, accepting and rejecting.

[0358] "Interest vector" refers to a multi-dimensional vector representation formed by the server after numerically modeling the user's interest characteristics based on behavioral history information. It is used to measure the degree of user preference on different categories of content.

[0359] A "preference vector" is a numerical vector used to represent a user's preference for a specific category, style, or attribute object. It is usually obtained by statistically analyzing historical behavioral information and performing feature extraction and normalization operations.

[0360] "Preprocessing information" refers to structured data obtained by the server after summarizing, cleaning, and transforming the original behavioral history information within a predetermined time interval, which is used for subsequent calculation of interest vectors or preference vectors.

[0361] "Candidate users" refers to multiple users selected by the server based on their behavioral history and emotional information who share common interests or are in similar emotional states, and are used for subsequent information exchange or product proposal processing.

[0362] "Messages" refer to text content addressed to one or more users to facilitate information exchange, interaction, or product proposals, including greetings, explanatory texts, reminders, etc.

[0363] "Topic information" refers to text or topic tags that can serve as the core of a conversation, guiding two or more users to engage in communication around that topic.

[0364] "Product proposals" refer to the recommendations and descriptions of products or services generated by the server based on users' historical behavior information, attribute information, and emotional information, in order to promote transactions or service selection.

[0365] In the following embodiments, the server, terminal, and user each act as the subject performing the corresponding operations. These embodiments are for illustrative purposes only and do not limit the technical scope of the invention.

[0366] I. Overall System Composition A server includes a processing unit, a storage unit, and a communication unit. The processing unit includes a central processing unit and an optional graphics processing unit. The server stores the operating system, application server programs, data processing programs, generative artificial intelligence model inference programs, and library files for natural language processing and sentiment analysis in its storage unit. The server exchanges data bidirectionally with multiple terminals via the communication unit.

[0367] The terminal includes a display device, an input device, a communication module, and a local storage device. The terminal can be a smartphone, tablet, personal computing device, or wearable display device. The programs executed by the terminal include a user interface program, a log collection program, and a client program for secure communication with the server.

[0368] Users input text, initiate browsing, click on recommended content, or respond to prompts using the terminal's input device. These actions are recorded as behavioral information on the terminal and sent to the server.

[0369] In one implementation, the server uses a general-purpose programming language as its primary implementation language. For natural language processing, the server utilizes open-source natural language processing libraries, such as those for word segmentation, part-of-speech tagging, and dependency parsing. For sentiment analysis, the server can use sentiment analysis libraries, such as sentiment analysis tools that combine dictionaries and machine learning. For structured data processing, the server uses data analysis libraries to convert behavioral logs from the database into table structures. For recommendation and vector modeling, the server uses machine learning frameworks, such as computing interest or preference vectors in a multi-dimensional feature space.

[0370] In some implementations, the server deploys a generative artificial intelligence model as an independent inference service. This generative AI model can be a multi-layer encoder-decoder network based on a transformer architecture, using a multi-head self-attention mechanism to encode the prompts and generate text word-by-word during the decoding phase. The server sends the prompt text to the model via an application programming interface (API) and receives the text results generated by the model.

[0371] II. Data Structure and Module Division The server maintains multiple data table structures for each user in the storage device. In the behavior log table, the server stores user behavior information in row records, with fields including user ID, timestamp, action type, object ID, dwell time, and interface ID. In the user attribute table, the server stores user attribute information, including age range tags, interest tag sets, and region tags. In the emotion record table, the server stores emotion information parsed from text, including emotion type, emotion intensity, and associated message identifiers.

[0372] The server stores the correspondence between generated data and the original behavioral history in the results table. This table records the generation time, prompt text, output text of the generative AI model, corresponding behavioral history summary, and sentiment tag. The server stores each user's interest vector or preference vector in the interest vector table. These vectors are represented as fixed-dimensional arrays of real numbers, with elements corresponding to predefined content categories or feature dimensions.

[0373] The server is logically divided into multiple functional modules. In the data acquisition and parsing module, the server receives behavioral information and text input reported by the terminal and performs syntactic analysis and normalization. In the state recognition module, the server uses rules and classification models to determine whether the user is in a state lacking information cues or a starting point for dialogue. In the prompt generation module, the server combines behavioral history, attribute information, and emotional information into structured features and dynamically generates prompts according to predefined templates. In the model invocation module, the server sends the prompts to the generative artificial intelligence model and receives the generated data. In the result management module, the server is responsible for storing and indexing the generated data. In the feedback update module, the server updates the interest vector and preference vector based on the user's reaction information and adjusts the future prompt generation strategy.

[0374] The server can use a supervised learning classifier in its state recognition module. The server can use a logistic regression classifier, support vector machine, or lightweight neural network to classify feature vectors. When constructing feature vectors, the server uses the following dimensions: dwell time on a particular session interface, the time interval between the most recent input, message length statistics, the number of clicks on recent recommendations, and the current interface type. When training the classifier, the server uses historical labeled data, in which users are manually or rule-basedly labeled to indicate whether they are in a state of silence or lack of informational clues. During the training phase, the server minimizes the cross-entropy loss function and updates the classifier parameters using gradient descent, enabling the model to distinguish between normal operations and operations requiring assistance within behavioral patterns.

[0375] III. Prompt Statement Generation and Model Calling The server generates different formatted prompts for different scenarios in the prompt generation module. In dialogue assistance scenarios, the server combines user interest tags, other party interest tags, recent chat content summaries, and current sentiment tags. The server uses natural language templates when constructing prompts, for example: "Based on the following information, generate three suitable opening lines for starting a chat with the user. User interests: {User interests}, User interests: {User interests}, Chat tone: relaxed and natural, avoid being too personal." In content recommendation scenarios, the server integrates user interest vectors, recently viewed content lists, and sentiment tags into prompts. For example: "Based on the user's viewing history and current mood, generate a list of recommended movies for the user to watch next, along with a brief reason for the recommendation. Viewing history: {historical data}, Current mood: {mood tag}." In product suggestion scenarios, the server writes the user's purchase history, current shelf category, and interest tags into the suggestion statement. For example: "Customer Interests: Movies, Past Purchase History: Multiple purchases of movie merchandise, Current Storefront: Latest movie merchandise. Please generate a natural product recommendation sentence, limited to 20 characters." In general information exchange scenarios, the server parses user-inputted task-oriented text into more explicit prompts. For example: "The user wants to share meeting PPT materials efficiently and securely with all group members. Please provide a detailed step-by-step solution and explain the tools used in each step." When generating prompts, the server doesn't simply concatenate text. Instead, it determines keywords or constraints in the prompts based on continuous features such as user interest vectors. The server can decide whether to add specific interests to the prompts based on whether a certain interest vector dimension value exceeds a threshold. For example, if action-related content has a higher weight, the prompt will explicitly specify "prefer action-related content".

[0376] The server sends the suggested text to the generative AI model within the model invocation module. When interacting with the generative AI model, the server can specify generation parameters such as maximum generation length, temperature parameters, and sampling strategies. The server obtains more stable recommended text under lower temperatures and richer candidate topics under higher temperatures.

[0377] In one embodiment, the generative AI model is a deep neural network based on a transformer architecture. The server incorporates multi-layered self-attention encoders in its model architecture, with input embeddings consisting of word vector sequences of prompts. In the decoder section, the server uses self-attention and encoder-decoder attention interactions to predict the conditional probability distribution of the next word. The server trains the model using extensive human dialogue data, recommendation corpora, and explanatory text. During training, the server uses cross-entropy as the loss function and employs an adaptive optimization algorithm to update the model parameters. During inference, the server freezes the model parameters and performs only forward computation.

[0378] IV. Modeling Interest Vectors and Preference Vectors The server periodically aggregates user behavior history in its data processing module. Within fixed time periods, the server calculates the weighted sum of each user's browsing frequency, dwell time, and click count for different content categories. The server constructs a user × category matrix using a data analysis library. Through normalization, the server maps each user's behavior values ​​for each category to a range of 0 to 1. The server uses this vector as an interest vector.

[0379] In another implementation, the server constructs the preference vector into higher-dimensional features, which may include content type, style tags, content length range, price range, etc. The server calculates a preference score for each dimension, allowing the preference vector to simultaneously represent preferences for multiple formats such as short videos, long videos, text and images, and audio.

[0380] The server embeds interest and preference vectors into the prompt statements when utilizing them. In one simple implementation, the server appends a natural language description to the prompt, such as "The user recently preferred action movies and science fiction movies." In another implementation, the server encodes the interest vectors into a list of tags, which the prompt model references during generation.

[0381] When selecting and ranking generated data, the server calculates the similarity between recommended candidates and user interest vectors. The server can use cosine similarity or dot product as the scoring function. When the generative AI model outputs multiple candidate descriptions, the server matches candidates with relevant content identifiers with the interest vectors, prioritizing candidates with high matching scores. This two-stage selection mechanism reduces potential bias in the generative model and improves overall recommendation accuracy.

[0382] V. Emotional Information Analysis and Candidate User Extraction After the user inputs text, the terminal sends the text to the server. The server performs word segmentation and syntactic analysis on the text in its natural language processing module. In its emotion recognition module, the server uses a hybrid approach combining dictionary-based emotion scoring with a supervised learning classifier. The server calculates an emotion polarity score and multidimensional emotion intensity, such as joy, sadness, anger, and tension, for each piece of text.

[0383] The server stores the sentiment vector for each message in a sentiment log table. In the candidate user extraction module, the server uses behavioral history information and sentiment vectors to find user groups with shared interests or similar current emotions. The server can use clustering algorithms to cluster users within the combined space of interest vectors and sentiment vectors, thereby selecting candidate users within the same cluster.

[0384] The server constructs prompt statements when generating message or topic information for candidate users. For example: "Based on the following information, generate a suitable topic to start a conversation between two users who share common interests and are currently in a similar mood. Common interests: {common interests}, current mood: {mood description}." After receiving the message information output by the generative artificial intelligence model, the server sends it to the corresponding terminal via a communication device. The terminal then displays the topic in the chat interface, allowing users to adopt or edit it with a single click.

[0385] VI. Technical Effects and Improvements in Computer Technology By employing the aforementioned structured data tables, feature vectors, and prompt generation mechanism, the server significantly improves the invocation process of generative AI models compared to traditional fixed-prompt methods. Instead of using static templates to generate prompts, the server generates different inputs based on dynamic behavioral history, interest vectors, and emotional information. This fully leverages the model's internal representational capabilities, enhancing the consistency between the output and the current state. This approach improves the efficiency of utilizing large-scale user behavior data within the server itself.

[0386] The server triggers topic or recommendation generation only when the user is in a stagnant state using a state recognition module, thereby reducing unnecessary model calls and lowering communication load and computational resource consumption. The server compresses high-dimensional behavioral data into fixed-dimensional vectors through pre-calculation of interest and preference vectors. This feature extraction significantly reduces the time complexity of subsequent similarity calculations, thus improving response speed.

[0387] In the emotion recognition and candidate user extraction process, the server improves matching accuracy and reduces the probability of erroneous recommendations by performing calculations in the vector space rather than simply relying on rule matching. During model training, the server enhances the model's robustness to diverse inputs by minimizing the loss function and using data augmentation techniques such as synonym substitution and word order perturbation. During inference, the server utilizes temperature and sampling strategies to adjust output diversity, achieving a balance between generation quality and diversity.

[0388] The server employs a unified data structure and modular pipeline throughout the end-to-end data flow, creating a closed-loop feedback mechanism between behavior logs, attribute information, sentiment information, and generated data. By recording user responses to the generated data, the server updates its interest vectors and generation strategies. This closed-loop feedback mechanism enables the entire system to automatically optimize during continuous operation, improving prediction accuracy and generation quality.

[0389] In this invention, the terminal serves not only as a display and input interface but also as a data collection node for behavioral data and real-time status, providing the server with a fine-grained data source. In wearable device scenarios such as smart glasses, the terminal reports the visual focus area and location data to the server. Based on this, the server triggers product proposals and topic prompts, directly linking the system of this invention to real-world physical locations and physical products, enabling device control and information overlay display in real-world scenarios.

[0390] Users do not need to manually write complex queries or search for information. Through natural operations and simple text input, the server performs behavioral pattern recognition, feature modeling, prompt construction, and generative artificial intelligence model invocation in the background. This process not only reduces the user's workload but also improves the overall processing efficiency and accuracy of the system by optimizing the computer's internal data structures and model invocation flow. It represents an improvement in computer technology itself, rather than simply automating manual operations.

[0391] VII. Other Implementation Methods and Alternative Structures In other implementations, the server can replace specific software components. For example, the server can use different natural language processing libraries, as long as they can provide word segmentation, part-of-speech tagging, and dependency analysis capabilities. The server can use different machine learning frameworks to calculate interest vectors and preference vectors, such as using linear models or matrix factorization methods instead of deep neural networks.

[0392] Servers can employ different network architectures for generative AI models, such as sequence-to-sequence models based on recurrent neural networks or text generation models based on convolutional structures. Servers can also use different loss functions, such as label-smoothed cross-entropy loss, to improve generation quality.

[0393] For interest modeling, the server can employ collaborative filtering algorithms to represent user behavior as a user-item matrix, and then decompose it to obtain latent semantic vectors, which are then written into an interest vector table. For state recognition, the server can use anomaly detection algorithms to identify sudden changes in behavioral patterns and determine whether auxiliary topic generation needs to be triggered.

[0394] In some implementations, the terminal can perform some sentiment analysis or behavioral feature calculations locally and send the results to the server to further reduce the server's computational load and network transmission volume. The terminal can also cache recently generated data locally for displaying prompts in offline scenarios.

[0395] In summary, the server, terminal, and user collaborate through specific data structures, specific prompt statement generation processes, and specific generative artificial intelligence model invocation methods in various embodiments of the present invention. This enables the system to achieve technical effects such as improved processing speed, enhanced recommendation accuracy, reduced error rate, and reduced communication load in various application scenarios such as dialogue assistance, content recommendation, and user matching, thereby demonstrating a substantial improvement to computer technology itself.

[0396] use Figure 14 The processing procedure is explained.

[0397] Step 1: Users perform operations on the terminal and generate raw input information.

[0398] Users can perform actions such as browsing content, clicking on items, entering text, staying in the chat interface, or viewing products on the terminal interface.

[0399] The input consists of the specific actions performed by the user on the terminal and the text content entered by the user in the input box.

[0400] The terminal calls the operating system's time interface and interface status interface locally to record each click, scroll, dwell time, page identifier, object identifier, and text input content as structured log entries.

[0401] The terminal packages these log entries into a message containing fields such as user identifier, timestamp, interface identifier, operation type, object identifier, dwell time, and text content.

[0402] The output is the raw input message prepared to be sent to the server.

[0403] Step 2: The terminal sends the original input information to the server and receives confirmation.

[0404] The terminal uses a secure transmission protocol via its communication module to send the raw input information message generated in step 1 to the server's receiving interface.

[0405] The input is the original input information message generated in step 1.

[0406] When sending a message, the terminal adds a session identifier and a retry flag to each message so that the server can control deduplication and retransmission in the event of network jitter.

[0407] The server parses the message format at the receiving end and verifies the integrity and validity of the user identifier and timestamp.

[0408] The output consists of the original input information that the server has received and written to the cache queue, as well as the acknowledgment response received by the terminal.

[0409] Step 3: The server writes the raw input information into the database and generates separate records for behavioral information, status information, and text information.

[0410] The server retrieves the original input information sequentially from the receiving queue, and writes the original message records into the behavior log table using the user identifier and timestamp as the primary key.

[0411] The input is a raw input message from the terminal.

[0412] Based on the operation type field, the server extracts data such as clicks, browsing, and dwell time into behavior information records, extracts the current interface identifier and dwell state into status information records, and extracts the text content field into independent text information records.

[0413] In the data processing program, the server standardizes the format of timestamps, converts the duration of stay into numerical values, and uniformly encodes object identifiers.

[0414] The output consists of normalized behavior history information, status information, and text information records stored in the database.

[0415] Step 4: The server determines whether the user is in a state of lacking information clues or a starting point for conversation.

[0416] The server reads recent records from the behavior history information and status information tables and combines them into a feature vector with time as the sequence.

[0417] The input is a set of historical behavior and status information of a user within a predetermined time window.

[0418] The server calculates the total time spent on a given screen, the time interval between the most recent valid input, the number of times the screen is opened, and the number of times the most recent recommendation has been adopted, and then standardizes these values ​​into numerical features.

[0419] The server inputs the feature vector into a state recognition model or a preset rule, such as determining whether the dwell time exceeds a threshold and there is no new text input, or the classification model outputs a "needs assistance" label.

[0420] The output is a flag indicating whether the user is currently in a state where there are no information clues or a starting point for a conversation.

[0421] Step 5: The server performs natural language processing on the text information and parses out the emotion information.

[0422] The server reads the most recently entered text content from the text information table.

[0423] The input is a string of the user's most recent text messages.

[0424] The server calls the natural language processing module to perform word segmentation, stop word removal, part-of-speech tagging, and syntactic analysis on the text, generating an intermediate representation containing words and part-of-speech tags.

[0425] The server uses an emotion dictionary and machine learning classifier in the emotion analysis module to score the sentiment tendency of each word, and then accumulates and normalizes the scores to obtain the overall emotion polarity and multidimensional emotion intensity.

[0426] The server writes the emotion type (such as joy, sadness, tension, etc.) and emotion intensity into the emotion record table in a structured form.

[0427] The output is a record of the emotion information corresponding to the text information.

[0428] Step 6: The server constructs user interest vectors and preference vectors based on behavioral history information, attribute information, and emotional information.

[0429] The server reads the user's browsing, clicking, viewing, purchasing records, and corresponding category tags during the predetermined period from the behavior history information table and the user attribute table.

[0430] The input consists of the user's historical behavior data set and attribute tag set.

[0431] The server counts the number of visits, weighted dwell time, and click-through rate for different categories of content, and normalizes the features of each category to generate a weight for each category.

[0432] The server arranges these category weights in a predefined order to form an interest vector, and combines the statistical results of various preference dimensions such as price range, content length, and style tags into a preference vector.

[0433] The server stores the latest interest vectors and preference vectors in the interest vector table and preference vector table, overwriting or appending historical versions.

[0434] The output is the user's current interest vector and preference vector.

[0435] Step 7: The server generates prompts for invoking generative artificial intelligence models.

[0436] The server reads relevant features from the interest vector table, preference vector table, sentiment record table, and the latest behavioral history information.

[0437] The input consists of interest vectors, preference vectors, sentiment information, and current contextual behavior information.

[0438] The server selects the corresponding prompt template based on the application scenario (dialogue assistance, content recommendation, product proposal, or task plan generation).

[0439] The server populates the prompt message template with user interest tags, other party interest tags, current mood description, recent behavior summary, etc., and dynamically adjusts the key constraints in the prompt message, such as "relaxed and natural", "limited to 20 characters", "generate 3 alternatives", etc.

[0440] For example, in a conversational assistance scenario, the server generates the following prompt text: "Based on the following information, please generate three suitable opening lines for the user to start a chat with the other party. User interests: {User interests}, Other party interests: {Other party interests}, Chat tone: Relaxed and natural, avoid being too personal." The output is one or more prompt text statements that can be used as input to a generative artificial intelligence model.

[0441] Step 8: The server calls the generative artificial intelligence model and obtains the generated data.

[0442] The server sends the prompt generated in step 7 as text input to the inference interface of the generative artificial intelligence model via the application programming interface.

[0443] The input consists of the prompt text and the generation control parameters (such as maximum length, temperature, and sampling strategy).

[0444] On the model side, the server encodes the prompt statement into a sequence of word vectors, which are then processed by a transformer encoder and decoder through multiple layers of self-attention operations. The probability distribution is calculated word by word, and the highest probability word is sampled or selected until an end marker is generated.

[0445] The server receives text information output by the generative artificial intelligence model, which may include multiple candidate dialogue openings, multiple recommendation result descriptions, or product proposal copy.

[0446] The output is a collection of the original generated data text corresponding to the prompt statement.

[0447] Step 9: The server evaluates, filters, and sorts the generated data.

[0448] The server will perform correlation calculations with the generated data and the user's interest vector, preference vector, and current sentiment information.

[0449] The input consists of a collection of generated data text, interest vectors, preference vectors, and sentiment information.

[0450] The server extracts tags or keywords related to content categories and topic types from the generated data and maps these tags into vector representations with the same dimension as the interest vector.

[0451] The server calculates the similarity between the generated candidate vector and the user's interest vector, for example, through cosine similarity or dot product operation, and can add sentiment matching degree as a weighting factor.

[0452] The server sorts the generated data based on similarity scores, removes candidates with similarity scores below a threshold, and retains only a few generated data entries that best match the user's interests and emotions.

[0453] The output is a list of filtered and sorted target data.

[0454] Step 10: The server will generate data and store it in association with behavioral history information, attribute information, and emotional information.

[0455] The server creates a result record for each piece of target-generated data. The record includes the generation time, the text of the prompt used, the generative artificial intelligence model identifier, the ranking score, and the associated behavioral history summary, interest vector, and sentiment label.

[0456] The input generates a list of data for the target, along with associated behavioral history, attribute information, and sentiment information.

[0457] The server writes the associated result records into the result table, using the user ID and generation time as index keys.

[0458] The server adds a version field to the records for subsequent analysis of the differences in the effects of different model versions or different prompting strategies.

[0459] The output is a set of generated data records that are persistently stored in the database.

[0460] Step 11: The server sends the generated data to the terminal and instructs the terminal to display it.

[0461] Based on the application scenario, the server encapsulates the target generated data into a response message, which includes text content, associated object identifiers (such as recommended content ID, candidate user ID, or product ID), and display style flags.

[0462] The input is a list of data generated for the target and the identifiers of the associated objects.

[0463] The server uses a security protocol via a communication device to send a response message to the corresponding user's terminal.

[0464] After receiving the response, the terminal generates corresponding UI elements on the interface based on the display style flags, such as recommendation cards, topic bubbles, or floating prompts.

[0465] The output consists of structured response messages that can be displayed on the terminal side, as well as prompts displayed on the interface.

[0466] Step 12: Users respond to the generated data on the terminal, and the terminal collects the response information.

[0467] Users can select, click, ignore, edit, or reject recommended content or topics provided by the server on the terminal interface.

[0468] The input consists of the generated data displayed on the interface and the user's interactive operations.

[0469] The terminal records the click action, dwell time, whether it was sent, whether it was deleted, and other feedback information for each piece of generated data, and associates this feedback information with the identifier of the corresponding generated data.

[0470] The terminal packages the response information into a feedback message, which includes the user identifier, the generated data identifier, the operation type, the operation time, and optional edited text.

[0471] The output is a response message that can be sent to the server.

[0472] Step 13: The server receives response information and updates behavioral history information and interest vectors.

[0473] The server receives the response information feedback message from the terminal and parses the generated data identifier and operation type within it.

[0474] The input is the feedback message of the reaction information uploaded by the terminal.

[0475] The server writes the reaction information into the behavior history information table and a special feedback table, and records behaviors such as "selected", "ignored", and "edited" as new behavior types.

[0476] The server updates interest and preference vectors based on feedback, for example, assigning higher weights to frequently adopted topics or content categories and gradually reducing the weights of categories that have been ignored for a long time.

[0477] The server uses the updated interest vector and preference vector when generating the next prompt statement and selecting the generated data, thereby achieving feedback-driven adaptive adjustment.

[0478] The output consists of updated behavioral history information, feedback records, and new interest and preference vectors.

[0479] 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.

[0480] 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.

[0481] 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.

[0482] 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.

[0483] 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.

[0484] Second Implementation Method Figure 3 An example of the configuration of the data processing system 210 according to the second embodiment is shown.

[0485] 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.

[0486] 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).

[0487] 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.

[0488] 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.

[0489] 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).

[0490] 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.

[0491] 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.

[0492] 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.

[0493] 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).

[0494] 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.

[0495] 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".

[0496] 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.

[0497] 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.

[0498] 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.

[0499] 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.

[0500] 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.

[0501] 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.

[0502] 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.

[0503] 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.

[0504] 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.

[0505] Third Implementation Method Figure 5 An example of the configuration of the data processing system 310 according to the third embodiment is shown.

[0506] 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.

[0507] 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).

[0508] 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.

[0509] 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.

[0510] 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).

[0511] 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.

[0512] 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.

[0513] 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.

[0514] 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.

[0515] 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.

[0516] 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".

[0517] 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.

[0518] 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.

[0519] 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.

[0520] 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.

[0521] 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.

[0522] 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.

[0523] 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.

[0524] 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 to analyze 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 to generate a menu using a generation AI. 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 to provide 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.

[0525] 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.

[0526] Fourth Implementation Method Figure 7 An example of the configuration of the data processing system 410 according to the fourth embodiment is shown.

[0527] 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.

[0528] 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).

[0529] 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.

[0530] 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.

[0531] 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).

[0532] 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.

[0533] 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.

[0534] 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.

[0535] 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.

[0536] 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.

[0537] 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.

[0538] 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".

[0539] 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.

[0540] 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.

[0541] 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.

[0542] 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.

[0543] 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.

[0544] 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.

[0545] 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.

[0546] 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.

[0547] 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.

[0548] 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.

[0549] 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.

[0550] 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.

[0551] 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).

[0552] 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.

[0553] 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."

[0554] 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.

[0555] 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).

[0556] 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.

[0557] 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.

[0558] 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.

[0559] 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.

[0560] 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.

[0561] 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.

[0562] 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.

[0563] 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.

[0564] 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.

[0565] 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.

[0566] In addition, the following notes are provided in response to the above explanation.

[0567] Example 1 (Note 1) An information processing system, characterized in that it comprises: A data processing device for receiving input information from a user terminal via a communication link and converting the input information into intermediate data for representing structured exchange information; A generation apparatus for generating prompt statements based on the intermediate data, the prompt statements being used to instruct a generative artificial intelligence model to generate exchange information for information exchange and / or a method for generating the exchange information; An acquisition device for inputting the prompt statement into the generative artificial intelligence model and acquiring response data including the exchange information and / or the generation method; A storage control device for converting the response data into a write request to the recording device and controlling the recording device to persistently store the response data; A forwarding control device for sending the response data instantly via a real-time communication link when the target user is in a connected state, based on the identification information of the target user corresponding to the response data and the communication status information of the target user, and for sending the response data when the target user reconnects when the target user is in a disconnected state; A historical management device for associating and managing the input information of the source user corresponding to the response data, the identification information of the source user, and the terminal identification information, thereby maintaining the associated information in the form of message history in the recording device; A status notification device for feeding back status information to the source user terminal, including the storage result of the response data and the forwarding result to the target user, and enabling the source user terminal to perform display control based on the status information.

[0568] (Note 2) The information processing system according to Appendix 1 is characterized in that, The data processing device is configured to parse multiple input information received from multiple user terminals and historical information stored in the recording device to generate prompt statements for the generative artificial intelligence model to generate abstract explanatory data representing common concerns among multiple users; the storage control device is configured to store the abstract explanatory data obtained from the generative artificial intelligence model as exchange information that can be shared by multiple target users in the recording device; the forwarding control device is configured to send the abstract explanatory data to the multiple target users.

[0569] (Note 3) The information processing system according to Appendix 1 is characterized in that, The data processing device is configured to analyze selected conditions obtained from users and behavioral history information stored in the recording device to detect recommended user groups with common interests and / or common attributes, and generate prompt statements for the generative artificial intelligence model to generate introductory messages and / or session initiation messages to facilitate information exchange among the recommended user groups; the storage control device is configured to store the introductory messages and / or session initiation messages output from the generative artificial intelligence model to the recording device; the forwarding control device is configured to send the introductory messages and / or session initiation messages to at least a portion of the users in the recommended user groups.

[0570] Application Example 1 (Note 1) An information processing system, characterized in that it comprises: A means for parsing input information from a user terminal and generating prompt statements based on the input information and the user's usage history to indicate the generation of conversation topics or recommendation information for information exchange; A means of inputting the prompt statement into a generative artificial intelligence model to generate information exchange data that is used as the conversation topic or the recommendation information; Means for obtaining the user's past usage history from a storage device, parsing the past usage history to extract the user's interest attributes, and reflecting the interest attributes in the prompt statement; A means of analyzing messages sent by the user and the dialogue history related to the messages through natural language processing to detect when a conversation is stagnant or lacks topic clues. Means for filtering the data for information exchange based on the user's interest attributes and the fact that the session is in a stagnant state, and for shaping the data for information exchange into candidate information for prompting the user; Means for parsing link information contained in information shared by the user, obtaining metadata from external resources corresponding to the link information, and converting the metadata into a form that can be displayed on the user's terminal; A means for storing the generated information exchange data and the metadata in the storage device and transmitting them to the target user terminal.

[0571] (Note 2) According to the information processing system described in Appendix 1, when it is detected that the conversation is in a stagnant state or the conversation lacks topic clues, the generative artificial intelligence model generates the conversation topic to be used as the initiation information for the continuation of the conversation by including the conversation history and the interest attribute in the prompt statement.

[0572] (Note 3) According to the information processing system described in Appendix 1, when parsing the link information, the system constructs the metadata from at least a portion of the title information, summary information, and image information obtained from the external resources corresponding to the link information, and displays the metadata in association with the information exchange data to the user terminal.

[0573] Example 2 (Note 1) An information processing system, characterized in that it comprises: Means for exchanging information between user terminals via communication processing devices; Means for obtaining the user's information sending history from a storage device based on user identification information; A means for parsing the sending time and content-related data contained in the acquired information sending history on a computing processing device in order to extract the user's behavioral patterns in information exchange. A means of determining, based on the behavioral pattern, current time, and terminal interface status information, whether the user has not sent a message within a predetermined time period and has not sent a message while the message input interface is open, thereby detecting the user's lack of information exchange clues. Means for generating prompt statements to instruct a generative artificial intelligence model to generate topic data that is presumed to be of interest to the user, based on the detection results and the behavioral patterns; Means for inputting the generated prompt statements into the generative artificial intelligence model and enabling the generative artificial intelligence model to generate topic data as clues for information exchange; This is a means of filtering and formatting generated topic data based on predetermined standards, and sending it to the user terminal as notification data. Means for recording the information sending history and the topic data into the storage device.

[0574] (Note 2) The information processing system according to Appendix 1 is characterized in that, The means for extracting behavioral patterns is configured to: input the information sending history into a behavior analysis artificial intelligence model built using a machine learning framework, and generate the behavioral patterns by calculating the feature quantities of user sending frequency, sending time period and topic tendency.

[0575] (Note 3) The information processing system according to Appendix 1 is characterized in that, The means for generating prompt statements is configured to insert interest categories, recent conversation topics, and types of expected conversation partners extracted from the behavioral patterns into a pre-defined text template, thereby forming prompt statements in natural language form that include the number of generated statements, expression length, tone, and prohibited topic conditions.

[0576] Application Example 2 (Note 1) An information processing system, characterized in that it comprises: A processing device for acquiring input information; A device used to determine, based on acquired behavioral or state information, whether a user is in a state where there are no information clues or a starting point for a conversation; An apparatus for parsing an information set containing a user’s behavioral history, attribute information, and emotional information, and for generating prompts that are input to a generative artificial intelligence model based on the information set. A means for inputting the prompt statement into the generative artificial intelligence model and obtaining generated data including at least one of information for information exchange, a topic as the starting point of a dialogue, or content recommendation information; A device for associating the generated data with the behavioral history information, attribute information, and emotional information and storing it in a storage device; A device for sending the generated data to a terminal device via a communication device and prompting it to the user; The device is used to obtain user reaction information from terminal devices, accumulate the reaction information as behavioral history information, and feed it back to the prompt statement and the device for generating and processing the generated data.

[0577] (Note 2) The information processing system according to Appendix 1 is characterized in that, The processing device is further configured to summarize and convert behavioral history information into preprocessing information at predetermined time intervals, calculate the user's interest vector or preference vector using the preprocessing information and include the vector in the prompt statement, evaluate multiple generated data output by the generative artificial intelligence model, and select or sort the generated data according to the interest vector or preference vector.

[0578] (Note 3) The information processing system according to Appendix 1 is characterized in that, The processing device is further configured to parse input information containing text information using natural language processing methods to determine emotional information, extract users with common interests or users in similar emotional states as candidate users based on the emotional information and the behavioral history information, generate prompt statements for instructing the generation of message information or topic information to promote information exchange or product proposals between the candidate users, input the prompt statements into the generative artificial intelligence model, and send the message information or topic information output by the generative artificial intelligence model to the terminal device corresponding to the candidate user.

Claims

1. An information processing system, characterized in that, include: processor; The processor is configured to: parse input information from the user, generate prompts to indicate the data required for information exchange; input the generated prompts into a generative artificial intelligence model to generate the data required for information exchange; and store the generated data in a recording device and forward it to the user as the recipient.

2. The information processing system according to claim 1, characterized in that, The processor is configured to: when generating information that can be shared among users, parse the information and use prompts to instruct the generation of data according to user needs, so as to generate data that meets user needs.

3. The information processing system according to claim 1, characterized in that, The processor is configured to: detect recommended users with shared interests in an environment that meets user-selected conditions, and use prompts to indicate the generation of information to generate information related to the recommended users.

Images