Information processing system

Abstract

The application provides an information processing system. An information processing system, characterized by comprising: a processor; wherein the processor is configured to: provide an interface for inputting user experience and user future target to a user; analyze the input information received by the interface, and generate prompt information for instructing a generative artificial intelligence model to perform specific processing; send the generated prompt information to the generative artificial intelligence model, and receive a response from the generative artificial intelligence model.

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] In existing technologies, when users engage in career planning or professional skills enhancement, they typically need to manually organize and express their experiences and future goals, and then manually search for relevant information or consult professionals. This approach has the following problems: First, users find it difficult to accurately and structurally express their experiences and future goals, leading to low efficiency in subsequent information acquisition and analysis. Second, existing general-purpose generative AI models are mostly for general dialogue or content generation purposes, lacking a mechanism to automatically construct adaptive prompts based on individualized user input, resulting in a discrepancy between model outputs and users' actual needs. Third, the responses obtained by users from generative AI models are often presented in raw text form, lacking visualization and highlighting of key points tailored to career development scenarios, making it difficult for users to extract valuable information for career development in a timely manner. Therefore, there is an urgent need for a system that can automatically generate adaptive prompts from users' input of experiences and future goals, and visualize the model responses in a way that is conducive to users' career formation and development, in order to improve the accuracy and efficiency of users' acquisition of career planning-related information. Summary of the Invention

[0004] To address the aforementioned issues, this invention provides an information processing system comprising a processor configured to: provide a user with an interface for inputting user experience and future goals, enabling the user to directly input their professional experience, current skills, and future aspirations in natural language; parse the input information received by the interface, performing word segmentation, semantic understanding, intent recognition, and key information extraction on the user input data using natural language processing technology, thereby generating prompts to instruct a generative artificial intelligence model to perform specific processing, and optimizing the prompts based on the parsing results to guide the generative artificial intelligence model to perform processing most closely matching the user's career development needs; send the generated prompts to the generative artificial intelligence model and receive responses from the generative artificial intelligence model; and visualize the responses of the generative artificial intelligence model as information related to the user's career development, including classifying, highlighting, and structuring the model response content, enabling the user to intuitively obtain key information related to career path planning, skills enhancement suggestions, and target job requirements. Through the above structure and processing flow, this invention enables users to obtain career development information that is highly relevant to their experience and future goals using a generative artificial intelligence model without needing professional prompt writing skills, thereby effectively improving the accuracy and usability of career planning support.

[0005] "System" refers to an overall device or assembly consisting of one or more hardware components and / or software modules for performing the series of processing steps described in this invention. The system may be located on a single physical device or distributed across multiple network-connected devices.

[0006] A processor is a computing unit that can execute program instructions, perform data operations, and control the coordinated operation of various functional modules of the system. It can be a general-purpose central processing unit (CPU), a graphics processing unit (GPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a combination thereof, or a processing module composed of multiple logic processing cores.

[0007] An "interface" refers to a combination of hardware and software provided by a system for interacting with users, including but not limited to graphical user interfaces (GUIs), command-line interfaces, web interfaces, and mobile application interfaces, used to receive user input and display output information to users.

[0008] "User experience" refers to career-related information accumulated by users in their past learning, work, or project activities, including but not limited to job experience, responsibilities, projects participated in, skills acquired, achievements, and related background descriptions. This information is usually entered by users in natural language.

[0009] "Future goals" refer to the state or plan that users hope to achieve in the future in terms of personal career development, skills improvement, or job transformation, including but not limited to the position they hope to hold, the technical fields they hope to master, the target industry, and career path planning, which are described by users in natural language through the interface.

[0010] "Input information" refers to data content related to user experience and future goals submitted by the user through the interface and received by the system. It can be a natural language description in text form, or it can include structured or semi-structured data.

[0011] "Parsing" refers to the analysis and processing process performed by the processor on the input information, including but not limited to word segmentation, part-of-speech tagging, syntactic analysis, semantic understanding, intent recognition, entity extraction, keyword extraction, and information structuring, in order to extract key content and semantic features suitable for subsequent processing from the input information.

[0012] "Prompt information" refers to the instructional text or data structure generated by the processor based on the parsed input information, used to guide the generative artificial intelligence model to perform a specific task. It usually includes task description, context information, constraints, and output requirements, and is used to clarify the processing goal and output form of the generative artificial intelligence model.

[0013] "Generative AI models" refer to AI models that can automatically generate text, code, images, or other content based on input prompts, especially language models or multimodal models built on deep learning and large-scale training data, used to answer users' career-related questions, generate suggestions, or create content.

[0014] "Specific processing" refers to the tasks related to the user's career development that the generative artificial intelligence model performs based on the prompts it receives, including but not limited to career advice generation, job requirement analysis, competency gap assessment, learning path planning, and resume optimization suggestions.

[0015] "Optimal processing" refers to a specific type of processing that best meets the user's career development needs and is most relevant and practical under the constraints of current user input information, model capabilities, and system strategies. It is usually determined by a comprehensive judgment of user intent, semantic features, and contextual information.

[0016] "Response" refers to the output of a generative artificial intelligence model after receiving prompts, including but not limited to natural language text, structured data, suggestion lists, analysis reports, or other information that can be further processed and displayed by the system.

[0017] "Career development" refers to the process of improvement and evolution of users in terms of job level, skill level, industry field, and scope of responsibilities during their career, including career planning, job selection, design of ability improvement paths, and career-related decision support.

[0018] "Visualization" refers to the process by which the processor processes the responses of generative artificial intelligence models into a presentation format suitable for human understanding and decision-making, and then displays it to the user through an interface. This includes, but is not limited to, displaying information in the form of text, lists, tables, charts, highlighted labels, structured cards, etc., so that users can intuitively understand and use information related to career development. Attached Figure Description

[0019] Figure 1 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the first embodiment.

[0020] 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.

[0021] Figure 3 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the second embodiment.

[0022] 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.

[0023] Figure 5 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the third embodiment.

[0024] 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.

[0025] Figure 7 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the fourth embodiment.

[0026] 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.

[0027] Figure 9 This represents an emotion map that maps multiple emotions.

[0028] Figure 10 This represents an emotion map that maps multiple emotions.

[0029] Figure 11 This is a sequence diagram illustrating the processing flow of the data processing system of the first embodiment.

[0030] Figure 12 This is a sequence diagram illustrating the processing flow of the data processing system in Application Example 1.

[0031] Figure 13 This is a sequence diagram illustrating the processing flow of the data processing system of the second embodiment.

[0032] Figure 14 This is a sequence diagram illustrating the processing flow of the data processing system in Application Example 2. Detailed Implementation

[0033] Hereinafter, an example of an implementation of the system according to the present disclosure will be described with reference to the accompanying drawings.

[0034] First, let me explain the terminology used in the following instructions.

[0035] 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.

[0036] 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.

[0037] 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.

[0038] 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.

[0039] 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.

[0040] First Implementation Method Figure 1 An example of the configuration of the data processing system 10 according to the first embodiment is shown.

[0041] 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.

[0042] 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).

[0043] 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.

[0044] 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.

[0045] 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.

[0046] 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.

[0047] Figure 2 The diagram shows an example of the main functions of the data processing device 12 and the smart device 14.

[0048] 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.

[0049] 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).

[0050] 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.

[0051] 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.

[0052] 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."

[0053] In existing career development guidance systems, users typically just forward their input of career experience and future goals to a generative AI model, which then directly generates suggestions. This approach presents several technical problems: First, the server-side lacks structured modeling and sophisticated natural language processing of user text data, resulting in a lack of specificity in the input passed to the generative AI model. This leads to unstable prompts, and the quality of the generated results fluctuates depending on the user's expression, resulting in poor overall system controllability and robustness. Second, the server-side lacks a mechanism for selecting prompt templates and automatically assembling them based on category information. It cannot adaptively generate clearly segmented and point-based prompts according to structured features such as career category, skill category, and years of work experience. This makes it difficult for the generative AI model to output results according to the expected paragraph and enumeration structures, thus reducing the parsability and usability of the output. Third, the server-side typically... The system directly presents the output of the generative AI model to the user without post-processing for computational processing. It lacks automatic extraction and differential calculation of professional ability information, learning project information, and action plan information in the response data. It cannot form a visual representation of structured information such as the difference between the current ability and the recommended ability, as well as the learning order along the time axis, thus limiting the quality of human-computer interaction and information presentation efficiency on the client side. Fourth, the server side does not record and update user input, prompts, and response data in a historical and feedback manner. It lacks a mechanism to dynamically optimize prompt templates and natural language processing parameters using historical information, making it difficult for the system to adaptively improve the overall generation quality and stability as the number of uses increases.

[0054] Therefore, a technical solution is needed to systematically improve the process of calling and utilizing generative artificial intelligence models on the server side through a series of computational steps, including standardized data verification, structured natural language processing, generation of prompt statement templates based on category information, structured parsing and visual editing of response data, and adaptive updates driven by historical information. This solution aims to improve the accuracy and uniformity of prompt statement construction, enhance the parsability and display efficiency of generated results, and strengthen the overall scalability and maintainability of the system, thereby achieving a computer technology improvement in the process of career development information processing.

[0055] 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.

[0056] In this invention, the server includes a unit for generating an input / output interface on a display device and receiving user career experience information and future goal information as text information; a unit for converting the text information into structured data and performing integrity and format verification according to a predetermined data format; a unit for performing natural language processing on the structured data, including word segmentation, word extraction, and attribute classification, to extract current ability information and target direction information, and selecting a prompt statement template from multiple prompt statement templates based on category information, and generating a prompt statement as input to a generative artificial intelligence model by inserting the text information and extraction results into the selected template; a unit for sending the prompt statement to the generative artificial intelligence model through a communication interface and receiving response data; a unit for extracting career ability information, learning project information, and action plan information from the response data, generating differential information representing the difference between the user's current ability and the recommended ability, and learning sequence information along the time axis, and editing it into display data for visualization; and a unit for recording the text information, prompt statements, and response data as historical information and updating the prompt statement template and natural language processing parameters based on the historical information. This allows for structured modeling and templated prompt generation of user career-related text on the server side, enabling stable and precise control over the input format and semantic range of the generative AI model. It also allows for structured parsing and differential calculation of the model's responses, thereby improving the automation and reusability of the prompt generation and response utilization process. This enhances the display structure and interactive experience of the generated results on the client side, achieving effective optimization of the computer technology level of the generative AI invocation process in career development counseling scenarios.

[0057] "Display device" refers to any information output device used to present graphical interfaces and text information to users, including but not limited to computer terminals, mobile terminals or other electronic devices with display functions.

[0058] "Input / output interface" refers to a graphical user interface that is presented on a display device, allowing users to input or edit information, and displaying processing results or feedback information.

[0059] "User career experience information" refers to textual descriptions of a user's past work content, job type, responsibilities, project experience, and skill application, etc., which are related to their career activities.

[0060] "Future Goal Information" refers to textual descriptions of a user's desired career direction, target position, expected skill level, development timeline, and other information related to their future career planning.

[0061] "Text information" refers to data information recorded in natural language, which can be represented by a sequence of characters and can be parsed and processed by computer programs.

[0062] "Data structure" refers to the internal data representation used to organize and represent textual information in computer memory or storage media, including but not limited to key-value pair structures, object structures, or record structures.

[0063] "Predefined data format" refers to the pre-defined data organization rules in the system regarding field types, length ranges, and whether fields are required or not, which are used to constrain and validate the format of input data.

[0064] "Structured data" refers to data that is organized and stored according to a predetermined data format, and has been divided into fields and types, so that various information is represented in the form of clear fields and attributes.

[0065] Natural Language Processing (NLP) refers to the technical process of automatically analyzing and processing natural language text, including but not limited to word segmentation, part-of-speech tagging, syntactic analysis, semantic analysis, entity recognition, and text classification.

[0066] "Word segmentation" refers to the process in natural language processing of dividing a continuous text string into several words or tags according to word boundaries.

[0067] "Word extraction" refers to the process of identifying and selecting keywords or phrases related to a predetermined task from text information.

[0068] "Attribute classification" refers to the process of classifying words or fragments extracted from text according to a predetermined category system in order to determine their attributes or types.

[0069] "Current competency information" refers to structured information derived from the analysis of a user's professional experience information, reflecting the user's current skills, knowledge level, years of experience, and job competency.

[0070] "Target direction information" refers to structured information derived from the analysis of users' future goal information, reflecting the user's desired career direction, target ability areas, and job type.

[0071] "Category information" refers to the labeling information extracted from structured data through natural language processing to describe the category to which the data belongs, including but not limited to occupational category, skill category, and work experience range.

[0072] A "prompt statement template" refers to a predefined text structure framework used to generate input for generative artificial intelligence models, which contains fixed text and placeholders for inserting variable content.

[0073] "Prompt statements" refer to natural language text that is formed by filling in user text information and extraction results into a prompt statement template, and is used to explicitly instruct generative artificial intelligence models on task content, context information, and output requirements.

[0074] "Generative AI models" refer to AI models that can automatically generate text output based on input prompts, including but not limited to deep learning-based language models or large-scale pre-trained models.

[0075] "Programming interface" refers to the calling interface used for data exchange between the server and the generative artificial intelligence model, including but not limited to application programming interfaces and network service interfaces.

[0076] A “communication protocol” refers to a set of communication rules used to transmit data between a server and a generative artificial intelligence model, including but not limited to network transport protocols and application layer protocols.

[0077] "Response data" refers to the output data generated and returned by a generative artificial intelligence model after receiving a prompt statement, which is usually represented in the form of natural language text.

[0078] "Occupational competency information" refers to structured information extracted from response data regarding occupational competency elements that should be possessed or improved.

[0079] "Learning Project Information" refers to structured information extracted from response data regarding learning units such as learning content, learning modules, course types, or practical projects.

[0080] "Action plan information" refers to structured planning information extracted from response data, which includes details about specific action steps, timelines, and execution sequences.

[0081] "Display data" refers to structured or semi-structured data that is suitable for visual presentation on a display device after the response data has been extracted, organized, and edited.

[0082] "Differential information" refers to the comparison result showing the difference between the user's current ability information and the recommended ability information in the response data, and is used to indicate the key ability points to be improved.

[0083] "Learning sequence information" refers to the sequence of learning items or learning stages arranged according to a timeline or logical progression, used to guide users to learn in sequence.

[0084] "List format" refers to a display format that presents multiple pieces of information in an itemized or numbered manner.

[0085] "Hierarchical format" refers to the display format that presents the hierarchical relationship of information in a tree structure, hierarchical structure, or indentation structure.

[0086] "Historical information" refers to the accumulated data recorded during system operation, including user text information, prompts, and response data, which is used for subsequent analysis and model optimization.

[0087] "Natural Language Processing Parameters" refers to configurable values ​​or thresholds used to control the behavior of the natural language processing process, including but not limited to word segmentation strategy parameters, classification thresholds, and weight coefficients.

[0088] The embodiments of the present invention will combine the collaborative work of servers, terminals and users in a real computing environment to specifically describe the system composition, data structure, algorithm flow and internal processing method of generative artificial intelligence model, so that those skilled in the art can implement the present invention accordingly.

[0089] I. Overall System Composition In one implementation, a server is deployed on a computing device in a data center. This device can be a general-purpose processor-based computer system, such as a rack server employing a multi-core central processing unit (CPU) and a graphics processing unit (GPU). The server operating system can be a UNIX-like system, such as a common server operating system, on which middleware and applications run.

[0090] In one embodiment, the server communicates with the terminal via web server software (e.g., an HTTP server software) and implements business logic through a backend application framework (e.g., a web framework based on a scripting language or a web framework based on an object-oriented language). The server can also communicate with generative artificial intelligence model services via an HTTP client library.

[0091] In one embodiment, the terminal can be a mobile terminal with a mobile operating system installed, a personal computer with a desktop operating system installed, or other electronic devices with display and input devices. The terminal generates a user interface using scripting languages ​​and front-end frameworks in a browser environment, or using native interface components in a local application environment.

[0092] Users input their professional experience and future goals through the input / output interface on the terminal. After the terminal performs preliminary verification of the input data, it sends it to the server. The server performs structured processing and natural language modeling on the data, generates prompts, calls generative artificial intelligence models, and finally returns the processed results to the terminal for display.

[0093] II. Server-side module composition and data structure In a preferred embodiment, the server includes the following functional modules, which are executed on the processor as software components: 1. Input data receiving and format verification module The server receives request messages from the terminal via a web interface. It uses a JSON parsing library to parse the text information in the request message into an internal data structure. The server then stores the user's professional experience and future goals as a record structure containing several fields, such as "experience_text", "goal_text", "industry", and "years".

[0094] The server performs predefined data format validation on the above fields, including: checking if the field exists, checking if the text length is within a threshold range, and checking if it contains unparseable control characters. The validation results are used by the logic judgment unit to generate Boolean flags and error codes for subsequent processing.

[0095] 2. Structured Data Generation and Feature Representation Module After successful validation, the server converts the text information into structured data. The server can map each user input to an internal "occupational record object," which contains multiple key-value pairs, such as occupational category, skill list, and experience range. The server can store this information in memory using a key-value mapping structure (such as a hash table) or an object-oriented data structure.

[0096] In one implementation, the server also constructs feature vectors for subsequent natural language processing and model invocation, mapping certain category information (such as occupational categories and skill categories) into discrete integer indices or sparse vector representations, which are used to select prompt statement templates and adjust generative artificial intelligence model parameters.

[0097] 3. Natural Language Processing and Category Information Extraction Module The server performs semantic analysis on career experience and future goal information using a natural language processing library. The server executes word segmentation algorithms on the CPU, such as a segmentation method combining dictionary matching and statistical models, to break continuous text into word sequences. The server can further utilize a pre-trained part-of-speech tagging model to tag each word with its part of speech for subsequent entity recognition and attribute classification.

[0098] In one embodiment, the server uses a deep learning-based text classification model to classify text segments. This text classification model can employ a bidirectional recurrent neural network, a convolutional neural network, or a lightweight language model based on an attention mechanism to encode the input word vector sequence and output labels such as occupation category, skill category, and experience level. Based on this, the server generates category information, including: - Occupational category tags, such as abstract categories like "software development," "data analysis," and "project management"; - Skill category tags, such as higher-level concepts like "programming languages," "databases," and "machine learning fundamentals"; - Work experience range labels, such as "0-2 years", "3-5 years", "5 years or more".

[0099] The above category information is written into a structured data object for subsequent prompt statement template selection and task constraints.

[0100] III. Algorithm for Selecting and Generating Prompt Statement Templates 1. Template Library and Index Structure The server pre-stores various prompt templates in non-volatile storage. Each template is associated with several category tags, such as matching a specific occupational category or skill direction. The server builds an index structure in memory to map category tags to template identifiers, thereby retrieving the appropriate template in O(1) or logarithmic time.

[0101] The template itself is a natural language text framework containing fixed text and several placeholders. Placeholders are used to insert the user's original text, extracted current capability information, and target direction information, etc.

[0102] 2. Template selection logic The server executes a template selection algorithm based on the extracted category information. For example, the server can use a weighted scoring rule: - If your job category is "Project Management" and your target area is "Artificial Intelligence Related Projects", then select the "AI Project Management Direction Template"; - If your job category is "Data Analysis" and your target direction is "Machine Learning + Business Analysis", then select the "Data Science Transition Template".

[0103] The server uses logical conditions and a weight matrix to calculate the matching score between each template and the current category information, and selects the template with the highest score. This process is executed on the server CPU, avoiding the terminal handling complex logic, thus unifying the template selection strategy.

[0104] 3. Prompt statement generation After selecting a template, the server inserts the user's professional experience information, future goals, and extracted structured key points into the template placeholder positions to generate prompts for the generative artificial intelligence model. The server concatenates the strings in memory, ensuring that the sentence structure conforms to the requirements of syntax and task instructions.

[0105] For example, when a user's scenario is "data analytics transformation to machine learning + business analytics", the server can generate the following example prompt: Current user status: - Job role: Data analysis related position; Years of work experience: 3 years; - Skills already acquired: Programming languages ​​and query languages; User's future goal: To work in a business analytics role that incorporates machine learning.

[0106] As a career development consultant and generative artificial intelligence expert, please answer using a clear point-by-point format based on the above information: 1. The additional machine learning and business analytics knowledge and skills that users need to acquire to achieve their goals; 2. Suggested learning sequence and practice methods (including course types and project practice examples); 3. Target job titles and corresponding core competency requirements that you can try within 1-2 years. In another example, the server can generate the following prompt for users with project management experience who wish to work on artificial intelligence projects: User Background: The user has been working in project management at a company for 5 years, responsible for the progress and quality management of multiple software development projects, and is familiar with agile development processes and Scrum management methods.

[0107] User goal: To enter the artificial intelligence industry in the future and work in AI project management or AI product-related positions.

[0108] Please analyze: 1. The user's existing advantages and transferability; 2. The knowledge and skills users need to acquire before entering the AI ​​industry; 3. Specific action plans that users can take in the next year (e.g., courses to take, types of projects to participate in, and networks to build).

[0109] Please provide specific suggestions based on current AI industry recruitment trends. Through the above template generation, the server can control the paragraph and enumeration structures of prompt statements while maintaining content flexibility, thereby improving the parsability of the output of generative artificial intelligence models.

[0110] IV. Structure and Invocation Methods of Generative Artificial Intelligence Models 1. Model Architecture In one implementation, the generative artificial intelligence model invoked by the server is a language model based on a deep neural network. This language model can employ a multi-layered self-attention transformer architecture, including: - Word embedding layer, used to map input word labels or sub-word labels into high-dimensional vectors; - Multi-head self-attention layer, used to calculate the relevance between words in different positions; - Feedforward network layer, used to perform non-linear transformation on the attention output; - Residual connections and layer normalization structures are used to stabilize the training and inference processes.

[0111] The model parameters are obtained through pre-training on a large-scale text corpus, employing either an autoregressive language modeling objective or a masked language modeling objective, and using the cross-entropy loss function as the error function. During training, the model updates its weight parameters via backpropagation, and the optimizer can use a gradient descent method with an adaptive learning rate.

[0112] 2. Model Reasoning Process The server sends prompts to the model service via a network communication interface. The model service performs forward propagation computation on the GPU, encodes the prompts into hidden representations, and progressively generates a sequence of output labels based on a probability distribution. During the generation process, temperature, top-k, or top-p sampling strategies can be used to control the diversity and determinism of the output.

[0113] After receiving the response text returned by the model, the server treats the response text as a natural language string for subsequent parsing and processing. By uniformly defining generation parameters (such as maximum output length and temperature value) on the server side, the length and diversity of generated results in different sessions can be kept within a controllable range, thereby improving system stability.

[0114] V. Response Data Analysis and Structured Visual Editing 1. Response Data Parsing Module The server performs further natural language processing on the response data returned by the generative artificial intelligence model, including segmentation, point recognition, and entity extraction. The server identifies professional competence items, learning items, and action steps in the response text using rule-based and lightweight classification models.

[0115] For example, the server can identify lines starting with "1.", "2.", "3." as first-level entries and lines starting with "-" or other markers as sub-entries. The server extracts and categorizes the skill names, time ranges, and action verbs in each entry to form structured lists of "professional ability information," "learning project information," and "action plan information."

[0116] 2. Generation of Differential Information and Learning Order The server performs differential calculations based on the user's current ability information and the recommended ability information in the response data. Through set operations or vector comparisons, the server identifies abilities that are not currently possessed but have been recommended multiple times and marks them as "abilities to be supplemented." The server assigns a weight or priority to each ability, which can be calculated based on the frequency of model suggestions, the relevance to the target direction, and the similarity to existing skills.

[0117] The server simultaneously generates learning sequence information based on time expressions (such as "Phase 1", "Next 3 Months", "Within One Year") and the order of entries in the response data, sorting the relevant learning items along a timeline. This sorting is achieved by maintaining a record structure in the server's memory that includes fields such as "Start Order" and "Estimated Duration".

[0118] 3. Display data generation and terminal presentation The server combines the aforementioned differential information and learning order information into display data. This display data can adopt a hierarchical structure, with the top level consisting of ability category nodes and the lower levels containing specific skill or learning item nodes. Each node includes attributes such as name, priority, and recommended time period.

[0119] After receiving the data, the terminal renders it into a list or tree structure. The terminal allows users to view their current skill gaps and planned learning stages hierarchically through interactive methods of collapsing and expanding nodes. This structured display moves beyond simply presenting long texts; it presents response results in a way that facilitates information understanding and management.

[0120] VI. Historical Information Recording and Adaptive Optimization 1. Historical information storage The server records user text information, generated prompts, and received responses for each session in non-volatile storage. The server generates a unique identifier for each record and attaches a timestamp and category label. This historical information is stored in a structured format for easy subsequent statistical analysis and model tuning.

[0121] 2. Template and Parameter Update Mechanism The server periodically analyzes historical information and statistically analyzes the performance of different prompt templates in terms of the quality of generated results. Quality is measured, for example, by user feedback scores, session duration, or the structured completeness of manually sampled evaluation results. Based on the statistical results, the server can adjust the usage weight of each template, modify fixed text prompts within templates, or add new templates.

[0122] The server can also adjust natural language processing parameters based on historical information, for example: - Adjust the word segmentation strategy to better adapt to occupational terms; - Adjust the classification threshold to reduce incorrect category predictions; - Update the skills dictionary to include names of emerging technologies.

[0123] Through the above feedback mechanism, the server implements an adaptive optimization process within the computer, enabling the system to continuously improve the accuracy of prompt generation and response parsing during use, thereby bringing about technical precision improvements and error reductions.

[0124] VII. Explanation of Technical Effects and Causal Relationship By transforming user natural language input into structured data, category information, and templated prompts, the server standardizes and constrains the input to generative artificial intelligence models, thereby reducing the model's sensitivity to differences in natural language expressions and improving the consistency and predictability of the generated results. This improvement directly enhances the stability of computational processing.

[0125] The server achieves secondary structuring of the response text through automatic calculation and visualization of differential information and learning order information, making the generated results easier for subsequent machine processing and human-computer interaction. This structuring process is not simply a human reading aid, but rather provides a more efficient data organization method for data management and retrieval within the computer, helping to improve query and display efficiency.

[0126] Through a historical information-driven template and parameter adaptive update mechanism, the server enables the system to continuously optimize the construction of prompt statements and natural language processing configuration without changing the underlying hardware. This results in a gradual decrease in error rate, an improvement in average generated quality, and an overall increase in computational efficiency in responding to user requests over long-term operation. This data-feedback-based self-optimization process constitutes an improvement to computer technology itself, rather than merely automating business processes.

[0127] VIII. Other Implementation Forms and Variations In another implementation, the server can avoid calling external model services and instead deploy a generative artificial intelligence model on a local GPU server. It can then use batch processing to perform unified reasoning on multiple prompts, thereby reducing network communication overhead and improving overall processing throughput.

[0128] In another implementation, the server can use category information not only for template selection, but also for dynamically controlling generation parameters. For example, it can reduce the generation temperature and limit the maximum output length for more specialized target directions, thereby further improving the rigor and relevance of the model output.

[0129] In one implementation, in addition to displaying text, the terminal can also draw visual graphics (such as capability radar charts and timeline charts) based on the displayed data. The generation of these graphics is based on the structured differential information and learning sequence information returned by the server, and is output in the form of images through the terminal's graphics rendering module, thereby expanding the application of the technical solution in information visualization processing.

[0130] The above-described embodiments can be combined or partially substituted for each other. Those skilled in the art can modify the module division, data structure type and algorithm details without departing from the spirit and essence of the present invention. All such modifications should be included within the scope of the technical concept of the present invention.

[0131] use Figure 11 The processing procedure is explained.

[0132] Step 1: The user opens the application interface on the terminal and enters information.

[0133] Input: Text of the user's professional experience, text of future goals, and optional information such as industry and years of work experience.

[0134] The terminal generates an input / output interface on the display device. Users can enter their professional experience and future goals in the corresponding text boxes via keyboard or touch screen, and select additional information such as industry and years of work experience in drop-down or radio button boxes.

[0135] The terminal reads the current value of each input control into a string or simple data type in memory, forming a set of raw input data for subsequent data encapsulation and processing.

[0136] Output: The set of raw user input data in the terminal memory.

[0137] Step 2: The terminal performs local validation on user input and encapsulates it into request data.

[0138] Input: The original user input data set obtained in step 1.

[0139] The terminal first checks whether the required text boxes are empty. If they are empty, it prompts the user to complete them on the interface. The terminal checks whether the length of each text field exceeds the preset limit. If it does, it prompts the user to shorten or truncate the text. The terminal also checks whether it contains obviously illegal characters (such as control characters).

[0140] After completing the verification, the terminal uses a local data structure (such as an object or key-value mapping) to organize fields such as "professional experience text", "future goal text", "industry" and "years of work experience" in a unified manner. Then, it serializes the data structure into structured data (such as a JSON string) for the request message body and caches the request to be sent in memory.

[0141] Output: Structured request data that meets the format requirements.

[0142] Step 3: The terminal sends structured request data to the server.

[0143] Input: The structured request data generated in step 2.

[0144] The terminal constructs an HTTP or HTTPS request message through the network communication module, writes content type information in the request header, and writes a structured request data string in the request body.

[0145] The terminal calls the operating system's network stack to encapsulate the request fragments into network data packets, which are then sent to the specified address and port of the server via the network interface.

[0146] Output: A network request message sent to the server port.

[0147] Step 4: The server receives the request and parses the raw text information.

[0148] Input: An HTTP / HTTPS request message from the terminal.

[0149] The web server software on the server receives requests on a designated port, parses the protocol header, and forwards the requests to the backend application. The backend application reads the structured request data string in the request body, calls a parsing library to deserialize it into an internal data structure object, and extracts career experience text, future goal text, and other fields from it.

[0150] The server performs field existence checks and basic format validations on the parsed results (such as reconfirming that fields are not empty and their length range), and generates corresponding status flags.

[0151] Output: A preliminarily validated object of user input data stored in the server's memory.

[0152] Step 5: The server converts text information into structured data and generates category features.

[0153] Input: The user input data object obtained in step 4 (containing original career experience text and future goal text).

[0154] The server calls string processing functions to perform text cleaning operations, including removing leading and trailing spaces, standardizing punctuation, and replacing invisible characters.

[0155] The server uses a natural language processing library to perform word segmentation and basic semantic analysis on the text of professional experience and future goals, and identifies tags such as potential occupational categories, skill categories and years of experience range based on a pre-trained text classification model.

[0156] The server writes these tags along with the original text into a "structured occupational record object," which represents fields such as "occupational category tag," "skill category list," "years of work experience tag," "original experience text," and "original target text," thus completing the mapping from free text to a structured representation.

[0157] Output: A structured occupation record object containing text fields and category feature fields.

[0158] Step 6: The server selects a prompt statement template based on the structured data.

[0159] Input: The structured occupational record object generated in step 5 (including occupational category, skill category, years of service, and other category information).

[0160] The server accesses a pre-built template index in memory and matches occupation category tags, skill category tags, and target direction information with the associated tags of each template in the template library.

[0161] The server performs a weighted scoring calculation, calculating a matching score for each candidate template based on factors such as tag matching, template applicability, and historical usage performance, and selecting one or more templates with the highest scores as candidates. In cases of identical matching scores, the server can also make a decision based on predefined priorities or random factors.

[0162] Output: One or more selected prompt template identifiers and corresponding template text frames.

[0163] Step 7: The server uses templates to generate prompts for generative artificial intelligence models.

[0164] Input: The prompt statement template selected in step 6 and the structured occupation record object obtained in step 5.

[0165] The server fills in the placeholder positions in the template text with the user's original professional experience text, future goal text, job category name, skill overview and work experience description, and performs string concatenation and line break formatting according to the predetermined sentence structure.

[0166] Through this data processing, the server recombines unstructured natural language input with extracted structured key points into prompts with clear paragraph structure and point lists, enabling generative artificial intelligence models to operate within a unified instruction framework.

[0167] For example, when a user has project management experience and wants to enter the field of AI project management, the server can generate the following prompt: User Background: The user has been working in project management at a company for 5 years, responsible for the progress and quality management of multiple software development projects, and is familiar with agile development processes and Scrum management methods.

[0168] User goal: To enter the artificial intelligence industry in the future and work in AI project management or AI product-related positions.

[0169] Please analyze: 1. The user's existing advantages and transferability; 2. The knowledge and skills users need to acquire before entering the AI ​​industry; 3. Specific action plans that users can take in the next year (e.g., courses to take, types of projects to participate in, and networks to build).

[0170] Please provide specific suggestions based on current AI industry recruitment trends. Output: A complete prompt text.

[0171] Step 8: The server sends prompts to the generative artificial intelligence model and retrieves response data.

[0172] Input: The prompt text generated in step 7.

[0173] The server constructs a request message for the generative artificial intelligence model service, writes the prompt text as an input field into the request body, and writes control parameters such as model name, maximum generation length, and temperature parameter.

[0174] The server calls the model service interface through a network client library, sending requests via HTTP or RPC to the inference server where the generative AI model is deployed. The inference server runs a language model based on a transformer architecture on its internal GPU, encodes the prompts, and generates a sequence of response texts token by token.

[0175] After receiving the response data returned by the model service, the server parses the response structure and extracts the response text in natural language form.

[0176] Output: A response text containing career advice.

[0177] Step 9: The server parses the response text and extracts structured suggestion information.

[0178] Input: The response text obtained in step 8.

[0179] The server again calls the natural language processing module to segment, divide, and recognize entities in the response text, identifying items that begin with a serial number or tag, and dividing the response into several suggestion units.

[0180] The server uses keywords and context rules to extract entities representing skill names, learning content, action verbs, and time descriptions from each suggested unit, and then categorizes and writes these entities into the "Professional Ability Information List," "Learning Project Information List," and "Action Plan Information List."

[0181] Through this data processing, the server transforms free-form long text responses into machine-processable structured lists, with each list element containing fields such as name, category, priority, or time attributes.

[0182] Output: A suggested dataset containing three types of structured information (professional skills information, learning project information, and action plan information).

[0183] Step 10: The server generates differential information and learning order information based on the current capabilities and the recommended capabilities.

[0184] Input: Current ability information from step 5 and recommended career ability information and learning project information from step 9.

[0185] The server performs set difference and intersection operations on the current capability list and the recommended capability list to identify the user's existing capabilities and missing target capabilities. The server calculates the priority for each missing capability, which can be calculated using a weighted formula based on indicators such as its position in the recommended list, its relevance to the target direction, and its similarity to existing capabilities.

[0186] The server simultaneously parses the time expressions and sequence descriptions in the learning project information, sorts the learning projects according to logical or chronological order, and generates learning order information. This information is stored in the form of a list grouped by stage, with each stage containing several recommended learning projects and estimated time periods.

[0187] Output: Differential information representing ability gaps and learning order information sorted by time or stage.

[0188] Step 11: The server generates display data for visualization and returns it to the terminal.

[0189] Input: The suggested dataset from step 9 and the difference information and learning order information from step 10.

[0190] The server integrates professional ability information, learning project information, action plan information, differential information, and learning sequence information to construct a hierarchical data structure. For example, "ability category" is the upper-level node and "specific skill or project" is the child node. The node attributes include tags such as "whether it is already possessed", "priority", and "recommended time period".

[0191] The server serializes the displayed data into a structured response message and sends it to the terminal via the network. Simultaneously, the server records the user input, prompts, and responses as historical information in its storage device for subsequent template optimization and parameter adjustment analysis.

[0192] Output: Structured display data sent to the terminal, and historical data stored persistently.

[0193] Step 12: The terminal receives and displays data and presents structured suggestions to the user.

[0194] Input: The structured display data returned by the server in step 11.

[0195] The terminal receives response messages through the network library and uses the parsing library to restore them into internal data structures, distinguishing different sections such as the capability difference list, the learning order list, and the action plan list.

[0196] The terminal generates corresponding interface elements on the display device according to the hierarchical structure, such as list controls, collapsible panels, or tree views, displaying content such as "existing skills," "skills to be improved," and "phased learning plan" in a partitioned format. Users can view detailed suggestions by scrolling, clicking, or expanding, thereby understanding their current skill level and the next learning path.

[0197] Output: A career development advice interface displayed on the terminal device, as well as visually structured content for users to interact with.

[0198] 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".

[0199] While existing computer information processing technologies possess systems capable of receiving user text input and invoking generative artificial intelligence (AI) models to generate relevant content, the following technical challenges remain: First, servers typically directly concatenate or paraphrase the user's raw input when generating prompts, lacking structured modeling and templated construction based on the user's long-term experience and target information. This makes it difficult to stably and controllably guide the model's output to highly match the user's goals, thus reducing the relevance and consistency of the generated results. Second, after obtaining the output of the generative AI model, servers often employ a one-off, unstructured display method, lacking the ability to fine-grained filter and reorganize based on keywords, paragraph structure, and user interest attributes. This results in a mixture of useful and redundant information, increasing the cognitive burden on the user at the terminal and hindering efficient acquisition of key information. Third, existing technologies often fail to systematically integrate user feedback and the generative AI model's generation history into the subsequent prompt generation process. Servers cannot dynamically adjust the prompt generation conditions based on the user's long-term behavior, making it difficult to achieve adaptive optimization of prompts and generated content at the system level.

[0200] Therefore, it is necessary to propose a new computer implementation scheme. By uniformly storing, analyzing, and modeling user experience information, target information, feedback information, and generation history information on the server side, and combining templated prompt statement generation, keyword and paragraph segmentation-based filtering, and user attribute-based scoring and selection strategies, the collaborative optimization of prompt statement generation and content filtering can be achieved within the computer. This will improve the processing efficiency and quality of generative artificial intelligence models in personalized information generation scenarios, thereby enhancing the performance and user experience of the entire information processing system.

[0201] 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.

[0202] In this invention, the server includes a processing unit for presenting a user interface on a communication terminal to obtain user experience information and target information as text information; a processing unit for converting the obtained text information into a storage data format and storing it in a storage device on a user-by-user basis; a processing unit for inserting the stored experience information and target information into a predefined template to form prompt statements as input to a generative artificial intelligence model; a processing unit for sending the formed prompt statements to the generative artificial intelligence model via a communication network and obtaining generated text data; and a processing unit for performing filtering processing on the generated text data based on keyword determination and paragraph segmentation, and according to evaluation rules associated with user interest attributes. The server includes a processing unit that extracts highly useful data to generate structured information data, and a processing unit that converts the structured information data into display data for visualization on the communication terminal display and distributes it to the communication terminal. Furthermore, the server also includes a processing unit for parsing user feedback and the generation history information of the generative artificial intelligence model when generating prompt statements, updating generation conditions within the template and adjusting the prompt statements sequentially, and a processing unit for dividing the generated text data into multiple information units during filtering, performing scoring calculations on each information unit based on user occupational domain information and target time information, and selecting information units based on the scoring results to form structured information data. This allows for collaborative optimization of prompt statement generation and content filtering within the computer through improved data structures and processing flows, enhancing the effectiveness and stability of generative artificial intelligence model invocation, reducing the occupation of irrelevant or redundant content on system resources and user terminal display, thereby improving the overall processing efficiency, response quality, and personalization capabilities of the information processing system, and achieving a substantial improvement in the information generation and distribution process based on generative artificial intelligence models in computer technology.

[0203] A "system" refers to a whole consisting of multiple communicating hardware devices and software programs running on them, used to perform a series of information processing functions such as user information collection, data storage, prompt statement generation, generative artificial intelligence model invocation, and result display.

[0204] "Information processing device" refers to an electronic device with a processor, storage device and communication interface, used to receive, store, process and control the output of input data, and is usually implemented in the form of a server device in this invention.

[0205] A "processing unit" refers to a logical functional entity in an information processing device, consisting of a processor and the software modules it executes, used to perform processing operations such as parsing, transforming, generating, filtering, and outputting input data according to a predetermined program.

[0206] "Communication terminal" refers to an electronic device that can send and receive data with a server through a communication network and has a user interface and display, including but not limited to mobile terminals, fixed terminals and other computing devices with network connectivity.

[0207] "User interface" refers to the set of interfaces or controls presented on a communication terminal for interacting with users. It is used to receive text information input by users and output visual information to users, including but not limited to interface elements such as input boxes, buttons, text areas, and lists.

[0208] "Experience information" refers to descriptive textual data about a user's past activities, work, study, projects, or other experiences, used to characterize the user's historical status and ability characteristics.

[0209] "Goal information" refers to descriptive text data about a user's desired future state, plan, development direction, or needs, used to characterize the user's expected goals and preferred directions.

[0210] “Text information” refers to digital information represented by sequences of natural language characters, including sentences, paragraphs, or other processable string data with characters as the basic unit.

[0211] "Storage device" refers to a hardware component for storing data, whether permanently or temporarily, including semiconductor memory, magnetic storage media, optical storage media, and combinations thereof.

[0212] "Recording medium" refers to the physical or logical medium used in storage devices to record and maintain digital data, including but not limited to disks, solid-state storage, and database files.

[0213] "Data format" refers to the data organization form or encoding structure specified for recording, transmitting and processing data, including field structure, encoding method and data type definition.

[0214] A "template" is a pre-defined string or data structure containing fixed parts and replaceable placeholder parts, used to insert user-related information into placeholders to form complete input content when generating prompt statements.

[0215] "Generative artificial intelligence models" refer to artificial intelligence models trained based on machine learning and deep learning techniques that can automatically generate output results such as natural language text based on input data, including but not limited to large-scale pre-trained language models.

[0216] "Prompt statements" refer to the instructional or guiding text used as input to generative artificial intelligence models. They are used to explicitly provide the model with context, roles, constraints, and output requirements, thereby controlling the type and form of the content generated by the model.

[0217] "Communication network" refers to wired or wireless network infrastructure used to transmit data between servers and communication terminals, including local area networks (LANs), wide area networks (WANs), and the Internet.

[0218] "Generated text data" refers to natural language text or equivalent parsable text output results automatically generated by generative artificial intelligence models based on prompts.

[0219] "Filtering" refers to the process of judging, screening, deleting, or reorganizing generated text data to remove irrelevant, redundant, or non-compliant content.

[0220] "Keyword determination" refers to the process of detecting whether a text contains specific keywords by using a pre-defined set of words or phrases, in order to evaluate the relevance of the text to user needs or a specific topic.

[0221] "Paragraph segmentation" refers to the process of dividing continuous text data into multiple logical paragraphs or information units based on line breaks, punctuation marks, bullet points, or preset rules.

[0222] "Evaluation rules" refer to the pre-set calculation standards or logical judgment conditions for evaluating the usefulness of text fragments or information units, including elements such as weights, thresholds, and scoring functions.

[0223] "User interest attributes" refer to a set of features such as user preference areas, topics of interest, and information styles inferred based on user input information, usage behavior, and feedback information.

[0224] "Structured information data" refers to data formed after text content is organized and labeled according to a predetermined data structure. It usually contains field-based information such as titles, body text, categories, tags or ratings, which facilitates subsequent processing and display.

[0225] "Display data" refers to output data that is organized and formatted for visual presentation on the display of a communication terminal, including content data that has undergone typesetting, segmentation, marking, and layout processing.

[0226] "Feedback information" refers to user interaction data such as ratings, selections, clicks, favorites, and marking as useful or useless after viewing generated content. It is used to reflect users' satisfaction and preferences with the generated content.

[0227] "Generation history information" refers to the data related to the generation process that is accumulated during the invocation of generative artificial intelligence models, including historical prompts, corresponding generation results, parameters used, and user feedback.

[0228] "Generation conditions" refer to the control parameters and constraints used when generating prompt statements or calling generative artificial intelligence models, including content scope, language style, output length, structural requirements, and preference settings related to user attributes.

[0229] An "information unit" refers to the smallest semantically independent content segment obtained from generated text data through paragraph division or other segmentation rules, which can be evaluated and selected individually.

[0230] "Occupational field information" refers to attribute information used to indicate the occupational category, industry type, or professional direction that a user is in or interested in.

[0231] "Target time information" refers to time attribute information related to the time range or stage that the user expects to achieve the target, including the target period, start and end time, or key time points.

[0232] "Scoring calculation" refers to the process of calculating numerical scores or grade evaluations for each information unit based on preset evaluation rules, which is used to measure the degree of matching between the information unit and user attributes and goals.

[0233] In one embodiment, the server includes: at least one multi-core central processing unit (CPU), an optional graphics processing unit (GPU), main memory, non-volatile storage, and a network interface controller. The server's software architecture includes: an operating system (e.g., a general-purpose server operating system), a database management system (e.g., a relational database management system), application server middleware (e.g., an HTTP-based application framework), and a generative artificial intelligence model inference module (e.g., a language model inference service deployed on the GPU). The terminal's hardware includes a processor, memory, a display, an input unit, and a wireless or wired communication module; its software includes a mobile operating system and a client application for executing the functions of this invention. Users interact with the server through a communication network between the terminal and the server.

[0234] In this embodiment, the information processing program running on the server is configured to: present a user interface on the terminal; receive user experience and target information from the terminal; store the received information in a database; generate prompts for a generative artificial intelligence model based on a predefined template; perform inference calls on the generative artificial intelligence model; obtain generated text data; perform filtering, scoring, and structuring processing on the generated text data; and convert the results into display data before sending it back to the terminal. Throughout the process, the server uses specific data structures to represent input data, prompts, generated text, and structured information to improve the locality of memory access and the efficiency of database queries.

[0235] In one implementation, the terminal displays two text input areas via a graphical user interface control, used for inputting "experience information" and "target information" respectively, and triggers data upload via a button. The terminal locally generates a data object containing the user identifier, experience text, and target text, and transmits this data object to the server via HTTPS. After receiving the structured information data returned by the server, the terminal maps each information unit to interface components, such as card views or list items, enabling the user to quickly browse the highly useful content filtered by the server.

[0236] In one implementation scenario, the user inputs the following natural language text as experience information and target information: "I have 5 years of experience in software project management, responsible for requirements communication and schedule control. I hope to apply generative artificial intelligence to project schedule prediction and risk warning." After receiving the text, the server marks it into two fields: experience information and target information, and stores them in corresponding data tables. For example, experience information is stored as a long text field describing the user's existing skills and project background, while target information is stored as a long text field describing future directions and needs. The server reuses these two fields in all subsequent generation and filtering processes, thereby avoiding repeatedly requesting user input every time the generative AI model is invoked, reducing communication load and user workload.

[0237] In one implementation, the server uses a relational database management system to maintain multiple tables on non-volatile storage. For example, the server maintains a user basic information table, a user experience target table, a generation history information table, and a user feedback information table. When storing user experience and target information, the server associates them with user identifiers and timestamps so that subsequent prompts can be generated based on historical data through multiple rounds of analysis. The server sets indexes for the user identifier and timestamp fields at the database level, thereby significantly reducing query latency when querying the latest records or sorting by time, thus improving processing speed.

[0238] The server uses a templated data structure when generating prompts. Multiple prompt templates are pre-stored in main memory, each containing a fixed text fragment and placeholders. After selecting a target template, the server fills the placeholder positions with the user's experience information and the target information. An example of a prompt is as follows: "You are a career development and artificial intelligence application consultant."

[0239] User's experience: 5 years of software project management experience, responsible for requirements communication and schedule control.

[0240] The user's goal is to apply generative artificial intelligence to project progress prediction and risk warning.

[0241] Based on the above information, please generate the following content in Simplified Chinese: 1. Explain the typical applications of generative artificial intelligence models in project schedule prediction; 2. Describe the application scenarios of generative artificial intelligence models in risk identification and early warning; 3. Provide suggestions for learning and practice steps over the next 6 months.

[0242] Please answer in sections, and provide specific, actionable suggestions. In another implementation, the server explicitly incorporates user interest attributes, professional domain information, and target time information into the template. For example, when the server infers a user preference for "case-based content" based on feedback, it adds a prompt to the template: "Please add at least three realistic case descriptions, highlighting practical details suitable for the software project management field." In this way, the server does not simply paraphrase user input but embeds control conditions related to the user profile within the prompt, thereby guiding attention to specific scenarios during internal model reasoning and improving the accuracy of the generated text matching user needs.

[0243] In one implementation, the server employs a neural network language model based on a self-attention mechanism as its generative artificial intelligence model. Structurally, this model comprises multiple layers of encoder-decoder stacks, each containing a multi-head self-attention sublayer and a feedforward network sublayer. Weights are obtained through pre-training on a large-scale corpus. The server fixes the model parameters during the inference phase and does not perform online retraining during user request processing. When invoking the model, the server encodes the prompt statement as a sequence of subwords, maps it to a vector sequence through an embedding layer, and then calculates the output vector sequence through several self-attention and feedforward layers. Finally, it obtains the probability distribution of the generated text through a lexical projection layer and a softmax function. When outputting the generated text, the server employs beam search or temperature-controlled sampling to achieve a balance between diversity and stability.

[0244] In another implementation, the server performs length control and segmentation encoding on the prompt statements before invoking the generative AI model. When a prompt statement exceeds a preset length threshold, the server splits it into multiple paragraphs according to semantic boundaries and connects them using paragraph markers, enabling the model to distinguish between different paragraphs during attention weight calculation. On the model's output side, the server truncates the text using a special end marker to prevent resource waste caused by infinite generation. Because the server controls text length and paragraph information at both the input and output ends, the computational complexity and memory usage during model inference are reduced, thereby improving overall processing speed.

[0245] After generating text data, the server employs specific filtering and scoring algorithms to technically process the text. First, the server performs segmentation, dividing the generated text into multiple information units based on line breaks, numbering symbols, and natural language sentence boundaries. The server then loads a set of user-related keywords, occupational information, and target time information from the database, using these as features to map each information unit into a vector representation. In one implementation, the server uses bag-of-words vectors or TF-IDF vectors, employing keyword frequency, position, and occupational-related terms as feature dimensions. In another implementation, the server uses a lightweight encoder model (e.g., a small self-attention network) to embed the information units to obtain semantic-level vector representations.

[0246] During the scoring phase, the server calculates a score for each information unit using pre-defined evaluation rules combined with user attributes. The server can employ a linear weighted scoring function, incorporating keyword matching, domain relevance, and time relevance as components, and deriving a comprehensive score through weighted summation. For example, for an information unit describing a "study plan for the next 6 months," the server increases its time relevance weight because the time interval is highly consistent with the user's target time information, thereby improving the comprehensive score. In another implementation, the server uses gradient boosting trees or logistic regression models to perform binary classification predictions on whether an information unit is "highly useful," using the predicted probability as the score. Through these algorithms, the server utilizes machine-executable mathematical operations rather than simple rule stacking, making the screening results more statistically stable.

[0247] The server performs filtering based on the scoring results, selecting information units for inclusion or exclusion. It sorts information units by score from highest to lowest, retaining those exceeding a threshold and assembling them into structured data according to a predefined structure, such as fields like "Concept Description," "Application Scenarios," and "Action Plan." During structuring, the server adds metadata tags to each information unit, such as source information, generation time, and keyword lists, for future traceability and reuse. This structuring and annotation not only reduces redundant text but also allows the terminal to employ a partitioned loading strategy during display, reducing the amount of data transmitted at once and thus decreasing communication load.

[0248] In one implementation, the server dynamically optimizes prompt statements using user feedback and generation history. The server records prompt statements, generates text data summaries, and user feedback tags (e.g., records of user clicks "helpful" or "not helpful") for each model call in the database. During periodic offline analysis, the server identifies which template configurations and generation conditions achieve high user satisfaction and then automatically adjusts the default parameters and content structure in the templates. In subsequent requests, the server adjusts the prompt statement generation conditions for the specific user based on their personal history, such as reinforcing content types preferred by the user and compressing parts that users generally ignore. Through this learning mechanism, the server achieves adaptive optimization of prompt statements and generated content at the system level, resulting in a simultaneous decrease in the average processing time of subsequent requests and the proportion of useless content.

[0249] In this invention, the server employs a combination of template generation, model inference, filtering and scoring, and structured processing, instead of simply passing user input through to the generative AI model and directly displaying the results. This unconventional processing sequence combined with a specific data structure brings several technical benefits: Firstly, through templated processing and user attribute injection, the model's attention is more focused on the feature subspace that the user cares about, thereby improving the relevance and text quality of the generated results. Secondly, by splitting the generated text into information units and performing scoring and selection, the server significantly reduces the amount of data that needs to be transmitted to the terminal, thus reducing the communication load. Thirdly, through structured information data and incremental display strategies, the server allows the terminal to load content on demand during rendering, reducing front-end memory usage and improving interface response speed.

[0250] In one implementation, the terminal uses a multi-region layout to display different information units based on the structured information data returned by the server. For example, the terminal fixes the "Concept Description" area at the top, and presents the "Case Studies" and "Action Plan" in a paginated format below. Based on field tags and rating results, the terminal prioritizes displaying high-rated content and provides an "Expand More" button for low-rated content, only requesting more data from the server when the user actively expands it. This lazy loading method further reduces actual network traffic, enabling faster response times even in mobile network environments.

[0251] In another implementation, users can provide feedback on various information units through interactive controls, such as marking them as "very useful," "average," or "useless." After receiving this fine-grained feedback, the server writes it into a user feedback information table and uses it as a supervisory signal in subsequent rating model training. The server can use these labeled samples to periodically update the rating model parameters to maximize user satisfaction-related indicators. This model update method based on real-world usage feedback, unlike traditional methods that rely solely on static rules, can gradually improve the system's performance for specific user groups over time, achieving a continuous improvement in the processing accuracy of the computer system, rather than simply replacing manual operations.

[0252] In a further implementation, the server can deploy multiple generative AI models of varying sizes and select between them based on factors such as the length of the prompt, the urgency of the user's needs, and the server's current load. For example, under high load, the server can prioritize models with smaller parameter sizes and faster inference speeds, while limiting the output length; under low load and when users require higher accuracy, it can use models with larger parameter sizes and better generation quality. When selecting models, the server quantifies the system state into feature vectors and makes decisions based on preset strategies or heuristic algorithms, thereby achieving a dynamic balance between processing latency and generation quality and improving the robustness of the entire system under different operating conditions.

[0253] In summary, by employing specific data structures (user experience target records, prompt statement templates, generation history, and feedback annotations), specific processing sequences (template construction, model inference, segmented scoring, structured output, and feedback loop), and control strategies optimized for computing and communication resources, the system provided by this invention achieves synergistic improvements in data management, model invocation efficiency, and output quality within the computer at the technical level. This differs from simple business process automation, achieving multiple technical benefits such as increased processing speed, improved accuracy, reduced communication load, and enhanced terminal display efficiency.

[0254] use Figure 12 The processing procedure is explained.

[0255] Step 1: Users input their experience and target information on the terminal.

[0256] In the application interface on the terminal, users enter text about their past work or study experiences in the "Experience" input box using the keyboard, enter text about their future goals in the "Goals" input box, and then click the "Submit" button.

[0257] Input: Raw natural language text strings (transferred text, target text) entered by the user in the terminal interface.

[0258] Output: Two text variables (experience information and target information) in the terminal memory, and a data object containing the user identifier and the above text.

[0259] Step 2: The terminal sends user input data to the server.

[0260] The terminal assembles the user identifier, experience information, and target information into a data object, serializes it into a string using JSON format, and sends the JSON as an HTTP request body to the server via the HTTPS protocol, calling a preset server interface URL. Before sending, the terminal performs UTF-8 encoding on the string and sets the content type and authentication information in the HTTP header.

[0261] Input: Data objects in terminal memory (user identifier, experience information, target information).

[0262] Output: An HTTP request message transmitted over the network to the server, where the request body is user input data in JSON format.

[0263] Step 3: The server receives and parses user input data.

[0264] The server receives HTTP requests from the terminal at the network interface, passes the requests to the application module, parses the HTTP headers and request body, reads the JSON string from it, and deserializes the JSON to obtain an internal data structure containing user identifiers, history information, and target information. During parsing, the server checks the integrity of the fields and removes leading and trailing whitespace.

[0265] Input: An HTTP request message (containing a JSON string) received from the network.

[0266] Output: Structured data objects (user ID, past text, target text) in server memory.

[0267] Step 4: The server stores user experience information and target information in the database.

[0268] The server checks the database for a corresponding record based on the parsed user identifier. If the record exists, an update operation is performed; otherwise, an insert operation is performed. The server converts the experience text and target text to an encoding format allowed by the database field (e.g., UTF-8) and appends a timestamp before writing it to the user experience target table. The server executes SQL statements through the database driver and returns the execution results upon completion.

[0269] Input: Structured data objects (user ID, past text, target text) in server memory.

[0270] Output: The updated data record in the database, and the write result status (success / failure, record ID) in the server memory.

[0271] Step 5: The server constructs prompts for generative artificial intelligence models.

[0272] The server reads the user's experience and target information from newly written or existing database records and selects a predefined prompt template. In memory, the server replaces placeholders in the template with the actual experience and target text, and can add additional constraint paragraphs to the template based on the user's professional field, interests, and target time information. The server generates the complete prompt text through string concatenation and replacement operations.

[0273] Input: Experience information and target information stored in the database, as well as prompt statement templates and user attribute data in the server's memory.

[0274] Output: The completed prompt text in the server's memory, used to send to the generative artificial intelligence model.

[0275] Step 6: The server sends prompts to the generative artificial intelligence model and retrieves the generated text data.

[0276] The server takes the prompt statement as one of the input parameters, along with model identifiers, maximum output length, temperature, and other parameters, and encapsulates it as API request data. This data is then used to call the generative AI model inference service interface via HTTPS. The generative AI model performs word segmentation, vector embedding, and multi-layer neural network forward propagation calculations on the prompt statement, outputting the corresponding natural language text. The server receives the JSON response returned by the model, parses the generated text field, and obtains the generated text data.

[0277] Input: The prompt text and the set of inference parameters in the server's memory.

[0278] Output: Generated text data (natural language strings generated by the model) in the server's memory.

[0279] Step 7: The server performs segmentation, keyword determination, and filtering on the generated text data.

[0280] The server first segments the generated text into multiple information units according to rules, based on markers such as line breaks, numbering symbols, and periods. Each information unit is treated as a relatively independent text segment. The server then loads a list of keywords and interest attributes relevant to the user, calculates the frequency and position of keywords for each information unit, and determines the relevance score to the user's professional field and target time. Based on preset evaluation rules, the server deletes or downgrades information units with low relevance or obvious redundancy, retaining only those with higher scores.

[0281] Inputs: generated text data in server memory, a list of user-related keywords, user's professional field information and target time information, and evaluation rules.

[0282] Output: A set of filtered information units in the server's memory, each with a relevance score and keyword hit information.

[0283] Step 8: The server performs scoring calculations on the filtered information units and generates structured information data.

[0284] The server performs a scoring operation on each information unit, combining features such as keyword matching, domain relevance, and time relevance according to preset weights to form a comprehensive score. Based on the comprehensive score, the server sorts the information units, selects several units with scores above a threshold, and organizes them into structured information data objects according to function (e.g., "Concept Description," "Application Scenarios," "Action Plan"). During the structuring process, the server generates a title or section description for each information unit and adds metadata (such as generation time and source information).

[0285] Input: The set of filtered information units and their feature values, scoring weight configuration, and structured rules.

[0286] Output: A structured information data object in the server's memory, containing multiple fully labeled information units.

[0287] Step 9: The server converts structured information data into display data and sends it to the terminal.

[0288] The server maps structured information data into a suitable data format for front-end display based on the terminal type and display capabilities. For example, it generates title, summary, and body fields for each information unit. The server can assign display priorities based on the scoring results so that the terminal can present the data according to the priority. The server serializes the display data into a JSON string and sends it to the terminal via an HTTP response.

[0289] Input: Structured information data objects in server memory, terminal display capability information.

[0290] Output: Display data sent over the network to the terminal (JSON format response message).

[0291] Step 10: The terminal parses and displays the data and presents it to the user.

[0292] After receiving the HTTP response from the server, the terminal parses the JSON string in the response body and maps the structured information within it to its internal data model. Based on fields such as title, summary, body, and priority, the terminal creates corresponding interface components for each information unit, placing high-priority content in a prominent position on the interface and providing scrolling or expansion entry points for other content. The terminal then uses local font settings and screen size for layout during display, providing users with a clear and readable content presentation.

[0293] Input: Display data (JSON format) received from the server, and local terminal interface configuration.

[0294] Output: Multiple text segments and corresponding interactive controls are displayed on the terminal screen for users to browse and operate.

[0295] Step 11: Users provide feedback on the displayed content, and the server updates the generation conditions based on the feedback.

[0296] In the terminal interface, users click feedback buttons such as "Helpful" or "Not Helpful" on various information units. The terminal converts these feedback actions into data objects containing user identifiers, information unit identifiers, and feedback types, and sends them to the server via HTTPS. After receiving the feedback, the server writes the feedback record to the user feedback information table and uses this feedback to update the generation conditions and scoring weights in the template during offline analysis or on the next request, making the subsequently generated prompts and filtering strategies more in line with the user's preferences.

[0297] Inputs: User feedback on the terminal, historical generation records stored on the server, and template configurations.

[0298] Output: Newly added feedback records in the database, as well as updated template parameters and scoring weights, used for the next round of prompt statement generation and filtering.

[0299] 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.

[0300] 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."

[0301] In previous talent recommendation and meeting appointment technologies, members within an organization were typically selected from a structured database based solely on pre-defined rules or simple keyword matching, and the results were returned to the user in list format. This approach suffers from the following technical problems: First, the server-side processing of user-inputted experience information, future goal information, and free text requests often remains at a superficial keyword retrieval level, lacking unified semantic feature modeling. This results in limited recommendation accuracy and difficulty adapting to complex and ever-changing user needs. Second, the recommendation logic and generative AI models operate independently. The server often submits the user's raw text directly to the generative AI model, lacking prompts constructed based on the joint features of the user and candidate members. This leads to a disconnect between the generated results and the backend retrieval and matching results, failing to fully utilize the respective advantages of machine learning and generative AI models. Third, existing systems generally lack a unified computational process encompassing the entire chain from "user needs analysis → candidate member retrieval and matching → generative AI description generation → automatic meeting schedule registration." Servers typically require multiple heterogeneous services to collaborate, resulting in loose interfaces, complex call sequences, increased resource consumption, prolonged response times, and decreased usability. Fourth, in the meeting arrangement stage, many solutions rely on manual switching between multiple systems and manual comparison of schedule data. They fail to integrate the member matching results and schedule management interface into a continuous calculation process on the same server, resulting in insufficient end-to-end automation.

[0302] Therefore, it is necessary to provide a new computer implementation method that enables the server to perform deep semantic analysis on user experience information, future goal information, and prompt statements within a unified system, generate unified feature quantities, combine them with data on organizational members and achievements, perform high-precision matching through machine learning models, and further construct an adapted prompt text input generative artificial intelligence model based on the matching results. This results in recommendations that are consistent with the backend matching results and are more interpretable. Finally, under the same server control flow, the system automatically calls the schedule management interface to complete the registration of meeting schedules, thereby improving recommendation accuracy, response efficiency, and the overall automation and intelligence level of the system. This substantially improves the processing performance and resource utilization of computers in this type of business scenario.

[0303] 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.

[0304] In this invention, the server includes: a module for providing an input / output interface for user experience information and future goal information via a processing unit on an information processing device; a module for parsing natural language text based on the user experience information, the future goal information, and user-input prompts, and generating feature quantities representing user needs; a module for generating retrieval conditions and obtaining member candidate data from an information storage device recording organizational members and related achievements based on the feature quantities; a module for preprocessing the member candidate data and the user's feature quantities to generate learning or inference input data including numerical vectors; and a module for using the input data to perform machine learning... The system comprises the following modules: a module for calculating the fit of each candidate member and generating a candidate member list for the user based on the fit; a module for generating prompt text for input to a generative AI model based on the user's prompts and the candidate member list, and sending the prompt text to the generative AI model to obtain response text containing recommendation reasons or explanatory information; a module for visualizing the recommendation results containing the candidate member list and the response text through the input / output interface; and a module for obtaining the schedule information of the members selected by the user and the user, calculating their free time using a schedule management program interface, and automatically registering the meeting schedule. This allows the server to achieve high-precision matching of user needs and member data through feature modeling and machine learning models within a unified computational process. Simultaneously, it drives the generative AI model to generate explanatory recommendations consistent with the matching results through prompt text constructed based on the candidate list, and automatically completes meeting schedule registration within the same processing chain. This reduces cross-system interactions and redundant calculations, lowers resource consumption and response time, and improves the overall processing efficiency and intelligence level of the computer system in talent recommendation and meeting arrangement scenarios.

[0305] "Information processing device" refers to a general-purpose or special-purpose computing device that has a processor and memory, and is used to execute computer programs to process input data and produce output results, including but not limited to servers, computer terminals and embedded devices.

[0306] A “processing unit” refers to a logical functional entity that runs on an information processing device to execute a predetermined data processing flow. It can be implemented by one or more processors, memory and program modules running on them.

[0307] "User experience information" refers to background information related to a user's profession, skills, projects, or tasks formed in past activities, including years of work experience, skills acquired, projects participated in, and roles assumed, which can be descriptive data entered by the user or obtained from a database.

[0308] "Future goal information" refers to information related to the user's expected future career direction, job role, skill level, or type of project participation, used to express the user's expectations for career development or ability improvement.

[0309] "Prompt statements" refer to text data that is input by the user in natural language to express specific needs, preferences, or constraints, and is used to construct prompt text for processing by generative artificial intelligence models or other models.

[0310] "Natural Language Text Parsing" refers to the technical process of performing word segmentation, syntactic analysis, semantic analysis, and keyword extraction on strings containing one or more natural language sentences in order to obtain a structured or semi-structured representation.

[0311] "Features" refer to numerical or symbolic forms that are extracted from raw data by an algorithm and used to represent the computable attributes of the data in a model, including but not limited to vectors, scalars, category labels, or combinations thereof.

[0312] "Members within an organization" refers to individuals who belong to the same organizational entity and perform a certain function or task within that organization, including formal employees, contract workers, or other personnel included in the organization's personnel management system.

[0313] "Outcomes" refer to the recordable results generated by members of an organization in the course of performing tasks or participating in projects, including but not limited to project reports, technical documents, source code, patent documents, review records, and performance data.

[0314] "Information storage device" refers to the storage medium and its management system used to store and manage data in a searchable manner, including relational databases, non-relational databases, file storage systems and distributed storage systems.

[0315] "Member candidate data" refers to a set of structured or semi-structured data obtained from an information storage device based on search criteria, corresponding to multiple members within an organization and their associated outputs, and is used for subsequent matching and recommendation processing.

[0316] "Preprocessing" refers to the process of cleaning, normalizing, encoding, imputing missing values, combining features, or reducing dimensionality of data before it is input into machine learning models or other algorithms.

[0317] A “numerical vector” is an ordered set of one or more numerical elements used to represent text, categories, or other high-level semantic information in a vector space, so as to facilitate computation in machine learning models.

[0318] "Input data for learning" refers to the set of feature data used as samples for supervised or unsupervised learning during the model training phase, including model input features and optional target labels.

[0319] "Inference input data" refers to the set of feature data input into the model after the model has been trained and is used to perform prediction, scoring, or recommendation processing.

[0320] A "machine learning model" refers to a mathematical model or computational structure that is learned from training data and is used to predict, classify, cluster, or score unknown data. These models include, but are not limited to, linear models, tree models, and neural network models.

[0321] A “neural network model” refers to a computational structure consisting of multiple neuron nodes with parameters and their connections. It is a machine learning model that maps input features through multiple layers of nonlinear transformations to achieve functions such as classification, regression, or representation learning.

[0322] "Fitness" refers to a quantitative indicator that represents the degree of matching between members within an organization and the needs of specific users. It can be expressed in numerical form, such as probability, rating, or ranking score.

[0323] The "member candidate list" refers to a collection of one or more members within an organization that are selected from the member candidate data based on indicators such as suitability and arranged in a certain order, and is presented to the user for selection.

[0324] "Generative artificial intelligence model" refers to an artificial intelligence model that can automatically generate text, speech, images or other content based on input prompts. In this invention, it mainly refers to a text generation model that generates recommendation reasons or explanatory information based on prompts.

[0325] "Prompt text" refers to natural language text automatically constructed by the system based on user experience information, future goal information, prompt statements, and member candidate lists, which is used as input for generative artificial intelligence models.

[0326] "Response text" refers to the natural language text output by the generative artificial intelligence model after receiving a prompt text, which in this invention usually includes reasons for recommendation or explanatory information.

[0327] "Recommendation results" refers to a comprehensive set of output information presented to users for decision-making, including a list of candidate members and explanatory information associated with each member.

[0328] "Input / output interface" refers to the interface form used to interact with users, including graphical user interface, web interface, mobile application interface or command line interface, which can be used to display information and receive user input.

[0329] "Schedule information" refers to data related to the time arrangements of users or members within an organization, including busy / idle status, existing meeting times, available time periods, and time preferences.

[0330] "Schedule management application interface" refers to the application programming interface used to access, query and modify schedule information, including local calendar system interface or network calendar service interface.

[0331] "Meeting schedule" refers to the time arrangement record automatically created or updated by the system in the schedule management program interface for arranging communication or meetings between users and selected members.

[0332] In one embodiment of the invention, the server, as a core component of the information processing device, is configured to run on one or more physical servers or cloud servers. The server may employ a general-purpose processor and memory architecture, such as a computer hardware architecture based on a multi-core central processing unit and main memory. An operating system, such as a Unix-like operating system, is installed on the server, and application services, middleware, and a database management system are deployed on it. The server can establish a secure communication connection with terminals via a wired or wireless network. The terminal may be an electronic device with a display device and input device, such as a smartphone, tablet computer, or personal computer. The user interacts with the server through the terminal.

[0333] In a preferred embodiment, the server deploys multiple functional modules implemented by processing units at the application layer. Each module runs as a software component or service process and executes corresponding instructions in the processor. Specifically, the server includes: a user interface management module, a text parsing and feature generation module, a member data retrieval module, a feature preprocessing and vectorization module, a matching model inference module, a generative artificial intelligence interaction module, a result presentation module, and a schedule coordination and meeting registration module, etc. The modules interact with each other through an internal bus, inter-process communication, or remote procedure calls, thereby forming a continuous data processing pipeline from user input to meeting registration.

[0334] In this embodiment, the user inputs their experience and future goal information in natural language text format via the terminal's display screen and touch input or keyboard input. A script program running locally on the terminal performs basic integrity checks on the input, such as checking for emptiness and length within a preset range. The terminal then encapsulates the user's experience information, future goal information, and prompts into a structured message and sends it to the server via an encrypted communication protocol. The terminal does not perform complex semantic calculations during this process; it only handles data collection and display, allowing complex calculations to be centralized on the server side, thus facilitating unified optimization of computational performance and model deployment.

[0335] In a specific example, the experience information a user enters in the terminal could be: "I have 3 years of backend development experience, mainly using Python and microservice architecture." The future goal information could be: "I hope to transition to an engineer working on generative artificial intelligence models within the next two years." The prompt could be: "Please recommend a few colleagues who are experienced in generative artificial intelligence models and recommendation systems and are willing to mentor others." After receiving the aforementioned text data, the text parsing and feature generation module executes natural language processing algorithms on its processing unit to perform word segmentation, part-of-speech tagging, entity recognition, and dependency parsing. The server can implement this function using a natural language processing library or framework, such as a parsing pipeline built on general-purpose natural language processing tools. The server then calls a pre-trained semantic representation model, such as a language model based on a transformer architecture, to map user experience information, future goal information, and prompts into fixed-length numerical vectors. The server can encode each text segment into a high-dimensional floating-point vector, for example, a vector with several hundred or several thousand dimensions. These numerical vectors, as features of user needs, are stored in memory and used in subsequent steps in conjunction with member data features.

[0336] While the server encodes user text, the member data retrieval module reads member data and associated deliverables data from an information storage device. This information storage device can be a relational database management system running on a server or network storage device. The server sends structured queries via database access drivers, using member, skill, and project tables as a foundation to jointly query all members that meet basic filtering criteria. For example, the server can construct query conditions based on keywords extracted from user text, such as "generative artificial intelligence model" and "recommendation system," and filter member skill tag fields. The server can also exclude members who have left the company or whose tutoring is not currently available, based on the member status field. The retrieved member candidate data is temporarily stored in the server's working memory as a record set.

[0337] The feature preprocessing and vectorization module then normalizes the candidate member data. The server removes redundant spaces, standardizes capitalization, and replaces special characters in the member's text fields (skill descriptions, project descriptions, etc.) to eliminate text noise. For numerical fields, such as years of work experience, number of projects participated in, and performance ratings, the server performs missing value imputation and standardization, mapping each field to a uniform numerical range. For text fields, the server uses the same semantic representation model or word vector model as the user's text to compress each member's relevant text into a high-dimensional numerical vector; the server then concatenates the text vector with the numerical feature vector to form a comprehensive vector representing the member's overall characteristics. The server arranges all members' comprehensive vectors into a matrix form for batch calculation in the matching model inference module.

[0338] In the matching model inference module, the server loads a pre-trained machine learning model. This model can employ a multi-layer feedforward neural network structure, including an input layer, several hidden layers, and an output layer. The input layer receives a combined vector of user and member features, for example, by concatenating or combining user and member feature vectors to form a unified input representation. The hidden layers can be several fully connected layers, each followed by a non-linear activation function, such as a rectified linear unit or other activation function, to improve the model's fitting ability. During the training phase, the server uses a labeled dataset, labeling each pair of historical user requests and members to indicate whether they match or the strength of the match, thus constructing a supervised learning task.

[0339] During training, the server uses loss functions, such as binary cross-entropy loss or mean squared error loss, to calculate the error between the predicted fit and the true label. The server computes the gradients of the weight parameters for each layer using the backpropagation algorithm and iteratively updates the model weights using optimization algorithms, such as stochastic gradient descent or its improved variants. The server can leverage parallel computing and batch processing techniques to improve the computational efficiency of the training process. After training is complete, the server persistently stores the model parameters, loads the model during the inference phase, and performs forward propagation computations in memory.

[0340] During online inference, the server inputs user features and the feature vectors of each candidate member into a trained neural network model. The model outputs a scalar fitness value, reflecting the degree of match between the member and the current user's needs. The server stores the fitness results of all candidate members as an array and sorts them to generate a candidate member list. The server can select the top few members with the highest fitness as recommendations based on a preset threshold or a set number. Through this matching mechanism based on multidimensional features and deep neural networks, the server can make better use of the high-dimensional correlation between user experience, future goals, and member achievements compared to simple rule-based matching schemes, thereby improving matching accuracy at the algorithm level.

[0341] To improve the interpretability of the system and the user's understanding of the recommendation results, the generative AI interaction module further integrates with the generative AI model. Based on the user's original prompts, user features, and a list of candidate members, the server constructs prompt text for inputting into the generative AI model. For example, the server can generate prompt text for each candidate member in the following format: "User Background: {User Experience Information}; User Future Goals: {User Future Goal Information}; User Prompt: {Prompt Statement}; Candidate Member Information: Name: {Name}, Department: {Department}, Main Skills: {Skill List}, Representative Projects: {Project Summary}. Please generate a concise recommendation for this candidate member in Simplified Chinese, explaining why they are suitable to help the user achieve goals related to the generative AI model." The server sends the constructed prompt text to the generative AI model. The generative AI model can be deployed on the same server or a dedicated computing node, and internally employs a language generation network based on a transformer architecture. This network is pre-trained on a large corpus and can optionally be fine-tuned on the organization's data to adapt to specific business scenarios. After the server provides the prompt text to the model, the model generates natural language response text sequentially. The server filters and truncates the response text, removing potentially security-incompatible content, and stores the remaining text as recommendation information for that member. By comprehensively analyzing the structured features of candidate members and the user's needs text, the generative AI model can generate more flexible, coherent, and targeted recommendation reasons than simple template text, helping users efficiently understand the recommendation basis.

[0342] The results presentation module combines the candidate member list with corresponding recommendation information into structured data and displays it visually through the terminal interface. For example, the terminal can display the name, department, core skills, summary of representative achievements, and the generated recommendation reason on each candidate member card. Users can scroll through the data, click to view detailed information, or make selections. This presentation method not only provides ranking results but also reduces the user's cognitive burden through rich explanatory information.

[0343] When a user selects one or more recommended members and submits a meeting request on the terminal, the server's schedule coordination and meeting registration module is activated. The server retrieves the user's and selected members' schedule information from the internal user account management system and calendar management system. The server calls the calendar service's busy / idle query function through the schedule management program interface to scan time slices within a certain future time window. The server can use a discrete time slice intersection calculation method to map the busy / idle markers of each participant onto a time grid, and quickly determine the common idle time period for all participants through bitwise AND operations. This algorithmic time intersection calculation, which is not based on simple manual comparison, can run efficiently on the server side, significantly reducing the combinatorial complexity that arises as the number of participants and the time range increase.

[0344] After finding one or more candidate time slots that match the preset duration and user time preferences, the server selects the final meeting time based on the earliest available time or other strategies, and sends an event creation request to the calendar service via the schedule management interface. The server fills in information such as the topic, participant identifiers, start and end times, and meeting room or online meeting link in the event. Once the calendar service successfully creates the event, the server stores the meeting schedule's identifier and detailed information in a local database and returns a key summary to the terminal for display. Users can view notifications such as "A meeting has been arranged for you and the recommended members: Time: [Date], [Time], Method: Online Meeting" in the terminal.

[0345] In this entire processing chain, the server not only performs traditional data retrieval and display operations, but also improves the system's computational efficiency and accuracy in multidimensional data matching, text understanding, recommendation explanation generation, and temporal intersection calculation through specific data structure and algorithm design. Specifically, the introduction of high-dimensional vectorized representations of features enables the server to process large batches of candidate members using matrix operations in vector space, rather than simply using string comparisons or simple conditional filtering, thereby significantly improving matching speed and robustness. Simultaneously, by restricting the input of the generative artificial intelligence model to the candidate set filtered by the machine learning model, the server reduces the range of candidates that the generative model needs to process, effectively reducing computational load and response latency. Furthermore, the server adopts a unified modular software architecture, allowing data to flow in a structured form between internal modules, thereby avoiding repeated parsing and redundant calculations, further improving the overall utilization of computing resources.

[0346] In different implementations, the server can employ different neural network structures. For example, the matching model can be constructed as a multi-layered network with an attention mechanism to highlight the degree of attention users pay to certain skills or achievements. Alternatively, a graph structure model can be used, modeling organizational or collaborative relationships as a graph structure, utilizing graph neural networks to calculate the positional features of members within the organizational network, and inputting these features in conjunction with user features into the matching model. These variations all technically enhance the granularity and discriminability of the matching results by expanding the feature space and improving the network structure.

[0347] In this invention, the terminal not only performs simple input / output functions but can also handle some preprocessing in other embodiments to reduce server load. For example, the terminal can locally perform character encoding conversion, language detection, or simple noise reduction, and compress and decompress data transmitted over the network to reduce communication bandwidth consumption. In some scenarios, the terminal can also cache some recommendation results or explanatory information so that users can still view historical recommendation records even when the network connection is unstable.

[0348] In summary, this invention integrates multiple technical modules on a server, including text parsing, feature vectorization, deep matching model inference, generative AI description generation, and schedule intersection algorithm calculation. These modules are combined with specific data structures and processes to enable the system to achieve higher accuracy and efficiency in talent recommendation and meeting arrangement scenarios within an organization. This solution does not merely automate manual screening and appointment scheduling; rather, it fundamentally enhances the computer system's ability to handle complex semantic matching and multi-party resource coordination problems by improving the computer's internal data representation, model structure, and processing flow.

[0349] use Figure 13 The processing procedure is explained.

[0350] Step 1: Users input information via the terminal. Users input experience information, future goals, and prompts into the terminal interface and then submit the input.

[0351] Input: Natural language text typed by the user in the input box, including "experience information", "future goal information", and "prompt statements".

[0352] Output: A text data object temporarily cached by the terminal.

[0353] The terminal performs length and non-empty checks on the input, performs preprocessing such as character encoding unification and removal of leading and trailing spaces on each input field, and then combines the three text fields with the user identifier into a set of structured data for subsequent transmission to the server.

[0354] Step 2: The terminal encapsulates the request and sends it to the server. The terminal serializes the structured data obtained in step 1 into a message and transmits it to the server over the network.

[0355] Input: Structured data containing user identifiers, experience information text, future goal information text, and prompt text.

[0356] Output: The network request data stream sent to the server, and the request status while waiting for the server's response.

[0357] The terminal converts the data into a unified character set format locally, checks the message size, sends a request to the specified interface address of the server through an encrypted communication protocol, and maintains a request identifier internally to match the server's returned result.

[0358] Step 3: The server receives the request and parses the user input. The server receives network requests from the terminal at a designated interface and parses the user data contained therein.

[0359] Input: A stream of network packet data sent by the terminal.

[0360] Output: User experience text, future target text, prompt text, and user identification variables generated in server memory.

[0361] The server first authenticates and verifies the request, checking whether the user identifier and session token are valid. Then, it deserializes the message, extracting the text and identifier fields and storing them in a data structure in memory, providing raw input for subsequent natural language processing and feature calculation.

[0362] Step 4: The server performs natural language parsing and feature generation on the user's text. The server performs natural language parsing on the user's experience information, future goal information, and prompts, and converts them into numerical features.

[0363] Input: Experience information text, future goal information text, and prompt statement text stored in memory.

[0364] Output: One or more feature vectors and a list of keywords representing user needs.

[0365] The server invokes the natural language processing module to segment, tag, and filter stop words for each text segment, generating a word sequence. Then, it uses a semantic representation model (such as a transformer-based encoding network) to encode each text segment, calculating a high-dimensional embedding vector. Simultaneously, the server uses a keyword extraction algorithm to identify skill words, domain words, and target-related words from the text, forming a keyword list. The server then constructs a unified feature object from these vectors and the keyword list, serving as input for subsequent retrieval and matching.

[0366] Step 5: The server retrieves candidate data of organization members based on feature values. The server uses keywords and semantic information from user characteristics to retrieve data on members and their outputs from the information storage device.

[0367] Input: User feature object, including a list of keywords and text vector representations.

[0368] Output: A set of candidate member records, each containing a member identifier, attribute information, and a summary of the deliverables.

[0369] The server constructs search conditions based on the keyword list, matches the skill tag field and project domain field with these keywords, and generates a structured query statement. The server sends a query instruction to the database management system, jointly reads records that meet the conditions from the member table and the output table, and assembles the multiple records retrieved into a member candidate data set, which is temporarily stored in memory for subsequent feature preprocessing and model inference.

[0370] Step 6: The server performs feature preprocessing and vectorization on the member candidate data. The server cleans, standardizes, and vectorizes the candidate member data to generate input features that can be used by the model.

[0371] Input: A collection of candidate member data, including text and numeric fields.

[0372] Output: The comprehensive feature vector matrix of each member.

[0373] The server performs noise reduction on text fields such as skill descriptions and project descriptions of members, such as removing redundant symbols, standardizing capitalization and formatting, and then calls a text encoding model to convert these texts into high-dimensional vectors. The server performs missing value imputation and standardization on numerical fields such as years of service and number of projects of members, mapping numerical values ​​of different proportions to a uniform scale. The server then concatenates the text vectors with the standardized numerical features to form a comprehensive feature vector for each member, and stacks all member feature vectors into a matrix for the next matching calculation.

[0374] Step 7: The server uses a matching model to calculate member suitability. The server uses a machine learning model to calculate the degree of matching between each member and the current user's needs.

[0375] Input: User feature vector and member feature vector matrix.

[0376] Output: A list of fit scores for each member.

[0377] The server constructs a combined vector containing user and member features for each member, and then inputs these combined vectors in batches into a pre-trained neural network model. During forward propagation, the neural network model performs multiple linear transformations and non-linear activations on the input vectors, ultimately producing a scalar as a fitness score at the output layer. The server collects the output scores of all members, forming a one-dimensional score array, and stores this array in association with the corresponding member identifier, providing the basis for generating a sorted list.

[0378] Step 8: The server generates a candidate list of members based on compatibility. The server sorts and filters the fit scores to generate a candidate list of members for users.

[0379] Input: An array of member fit scores and a list of member identifiers.

[0380] Output: A list of candidate members sorted by fit, with each entry containing basic member information.

[0381] The server sorts the score array in descending order and rearranges the member identifiers in the same order, selecting the members with the highest scores as candidates. The server extracts the names, affiliations, main skills, and representative achievements of these members from the member candidate data set, and merges this information into a structured member candidate list, which serves as the main part of the recommendation result.

[0382] Step 9: The server uses text to prompt the construction of generative artificial intelligence models. Based on the user's prompts and the list of candidate members, the server generates prompt text for each candidate member to be input into the generative artificial intelligence model.

[0383] Input: Original prompt text, user experience and future goal text, member candidate list.

[0384] Output: A set of text prompts for each candidate member.

[0385] The server extracts a concise summary of user experience and future goals, concatenates this summary with the original prompt, and forms a user background paragraph. The server then retrieves a summary of each member's attributes and deliverables from the candidate member list and combines these with the user background paragraph to create a complete natural language description. The server generates a separate prompt text for each member, containing explicit instructions, such as "Please generate a concise recommendation for this candidate member in Simplified Chinese, explaining why they are suitable to help this user achieve goals related to the generative AI model," thereby guiding the generative AI model to output the required type of response.

[0386] Step 10: The server invokes a generative artificial intelligence model to generate recommendation text. The server sends the prompt constructed in step 9 as text to the generative artificial intelligence model and obtains the response text.

[0387] Input: A set of text prompts for each candidate member.

[0388] Output: A set of recommendation response texts corresponding to each candidate member.

[0389] The server sequentially or in parallel inputs text prompts into a deployed generative AI model service, requesting the generation of a natural language response. Internally, the generative AI model predicts and outputs a text sequence token by token based on the prompts. The server receives the generated text, performs length truncation and sensitive word filtering, and retains the remaining portion as the member's recommendation description. The server associates the description text with the corresponding member identifier, forming a "member-description" pair to enhance the interpretability of the recommendation results.

[0390] Step 11: The server provides assembly recommendations and sends them to the terminal. The server combines the candidate member list with the corresponding recommendation descriptions to form a complete recommendation result, and then sends it to the terminal.

[0391] Input: A list of candidate members and a set of response texts with recommendation descriptions.

[0392] Output: The recommended result data structure sent to the terminal.

[0393] The server internally constructs a result entry for each member, which includes basic member information, fit score, and generative explanatory text. Then, all entries are encapsulated into a unified result object. The server serializes this result object and sends it to the requesting terminal through a network interface, thus realizing the transmission of recommendation information from the server to the terminal.

[0394] Step 12: The terminal displays the recommendation results, and the user selects the target member. The terminal receives the recommended results returned by the server and displays them to the user in the form of a graphical interface, so that the user can make a selection.

[0395] Input: The recommendation result data structure sent by the server.

[0396] Output: A list of candidate members and the user's selection results displayed on the terminal interface.

[0397] The terminal deserializes and formats the received data, generating a list or card view on the interface to display the name, basic attributes, and descriptive text of each member. Users can select one or more members they wish to meet by clicking or checking boxes. The terminal organizes the user's selections into a list, caches it internally, and prepares to send it to the server for meeting arrangements.

[0398] Step 13: The terminal submits a meeting request to the server. The terminal sends the user's selected member information and time preference as a meeting request to the server.

[0399] Input: A list of member identifiers selected by the user and optional time range preference information.

[0400] Output: A meeting request message sent to the server.

[0401] The terminal reads the candidate member selection results confirmed by the user on the interface and allows the user to input or select the desired meeting time period, such as specifying the available time range within a certain week; the terminal combines the member identifier list with the time preference into structured data and sends it to the server through the network interface to trigger the meeting scheduling process.

[0402] Step 14: The server automatically calculates the meeting time and registers the meeting schedule based on the schedule information. The server calculates shared free time based on the user's and selected members' schedule information and registers the meeting in the scheduling system.

[0403] Input: The list of members and time preferences in the meeting request, and the schedule information of each participant obtained from the schedule management interface.

[0404] Output: Specific meeting time and meeting schedule record identifier.

[0405] The server uses a schedule management interface to query the busy / idle status of users and selected members within a predetermined time window. It discretizes the timeline into fixed-length time slices and constructs a busy / idle marker sequence for each participant. The server performs bitwise logical operations on each participant's busy / idle sequence to identify continuous intervals where all participants are free, and then filters suitable intervals based on user time preferences. The server selects one or more time slots as candidate meeting times, ultimately determining a specific time interval. It then creates a meeting event through the schedule management interface, writing participant and time information to an external calendar system. After successful event creation, the server obtains an identifier for the meeting event, which is used for subsequent queries and notifications.

[0406] Step 15: The server returns meeting information, and the terminal displays the meeting results. The server returns the registered meeting schedule information to the terminal, which then displays it to the user.

[0407] Input: Meeting details, such as the meeting event identifier and time, returned by the external calendar system.

[0408] Output: Meeting arrangement information displayed on the terminal interface.

[0409] The server extracts the meeting start time, end time, meeting link or location, and participant list from the calendar system response, combines them into a meeting result object, and sends it to the terminal. The terminal receives the object and displays it in text form on the interface, such as "A meeting has been arranged for you and the selected members: the time is [date] and the format is online meeting." It also provides operation options such as "add to local calendar" or "copy meeting link," allowing users to view and manage the meeting later through the terminal.

[0410] 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".

[0411] In existing technologies, most systems for processing user experience, future goals, and booking intentions simply parse natural language input into a few fields and then perform static rule matching or fixed process scheduling. This presents the following technical problems: First, the system struggles to achieve deep semantic understanding of complex natural language input and cannot dynamically generate highly adaptive machine processing instructions based on context and subtle differences. This results in the underutilization of generative AI models, low computational resource utilization efficiency, and unstable recommendation results. Second, existing systems typically separate "recommendation processing" from "schedule setting processing." Recommendation results are not uniformly modeled with specific available time, workload, and contextual constraints within the same computational process. This leads to multiple manual interventions required between recommending individuals and scheduling, resulting in low automation. The overall processing latency is high; third, existing systems make recommendations based on rules or simple emotion tags, without incorporating the user's emotional state as a primary computational input into the complete chain of "prompt statement construction - generative AI model invocation - schedule calculation". This results in the system being unable to adaptively adjust the content of prompt statements and appointment time slots according to real-time emotional states, leading to significant deficiencies in human-computer interaction experience and intervention timing selection; fourth, in traditional systems, the generative AI model is treated as an isolated text generation module. The input and output lack a tightly coupled data flow design with the backend data structure and external schedule management services. The model output often requires manual secondary processing to drive subsequent schedule or resource allocation updates, making it difficult to form an end-to-end automatically closed-loop computational path.

[0412] Therefore, it is necessary to provide a new system architecture and processing mechanism to improve the collaborative methods of natural language input parsing, automatic generation of prompts, generative artificial intelligence model invocation, and external schedule management services from the perspective of computer technology. Furthermore, it is necessary to introduce emotion state estimation results into the same program execution chain to form adaptive control over recommendation results and schedule setting results, so as to improve the automation level, resource scheduling efficiency, and human-computer interaction quality of the overall system.

[0413] 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.

[0414] In this invention, the server includes: means for providing a user information input interface configured to receive natural language input from the user regarding experience, future goals, and schedule intentions; means for performing language processing on the natural language input obtained from the user, the language processing being configured to extract candidate individuals or task subjects from the constituent elements within the organization based on the user's experience and future goals, and generate prompt statements that instruct a generative artificial intelligence model to perform recommendations for individuals or task subjects and to assign or suggest tasks, together with additional information including the user's emotional state and work status; and means for sending the prompt statements to the generative artificial intelligence model and obtaining information from the generative artificial intelligence model including recommendations. The device includes a response containing a recommended candidate or suggestion; a scheduling device for acquiring behavioral schedule information of organizational components corresponding to the recommended candidate included in the response, determining the free time of both parties based on the user's desired schedule and the behavioral schedule information, and automatically scheduling interviews or job support during the free time; a device for inferring the user's emotional state based on the user's natural language input and the response from a generative artificial intelligence model, and adjusting at least one of the content of the prompt statement and the schedule time period set by the scheduling device based on the emotional state; and a device for providing a display interface configured to visually output the response from the generative artificial intelligence model and schedule-related information set by the scheduling device. This allows for the tight coupling of natural language parsing, emotion state estimation, dynamic construction of prompts, invocation of generative artificial intelligence models, and schedule updates from external schedule management services within the same server-side computation process. This enables integrated automatic processing of person-oriented recommendations and schedule settings, reducing human intervention, improving the mapping efficiency between generative artificial intelligence model outputs and structured scheduling data, and adaptively adjusting prompts and schedule time periods through emotion states. Consequently, it significantly enhances the intelligence and response quality of human-computer interaction while improving the task allocation and resource scheduling performance of the computer system.

[0415] "User" refers to an individual or group that provides information such as experience, future goals and schedule intentions through the user information input interface of the system and receives recommendation results and schedule setting results.

[0416] "Experience" refers to the knowledge, skills, project experience, and related achievements that users have gained in past activities, which can be used by the system as a basis for recommendations and schedule settings.

[0417] "Future goals" refers to the user's career development goals, learning directions, project participation intentions, or other long-term plans that they hope to achieve in the future.

[0418] "Schedule preferences" refers to users' preferences related to time management, such as the date, time period, duration, and format of interviews or assignment support.

[0419] "Natural language input" refers to unstructured text or voice content provided by users in a manner similar to everyday communication, which needs to be converted into structured data through language processing.

[0420] "User information input interface" refers to an interactive interface presented on a terminal or other computing device for receiving natural language input from users, including text input boxes, voice input controls, and related interactive elements.

[0421] "Language processing" refers to the parsing operations performed on natural language input, including word segmentation, part-of-speech tagging, entity recognition, intent recognition, keyword extraction, and semantic analysis.

[0422] "Organizational constituents" refers to entities within an organization that are managed or identified by the system, including people, work subjects, functional units, or resource units that can be assigned tasks.

[0423] "Candidate" refers to an individual within the set of personnel identified as potentially having the opportunity to interview, coach, or collaborate with a user within the organization's constituent elements.

[0424] "Working entity" refers to the entity that performs a specific operation, task, or process, including human workers, automated equipment, or a combination of both.

[0425] "Emotional state" refers to psychological and emotional characteristics inferred based on user input data, including quantifiable or categorizable states such as stress level, positivity level, anxiety level, and expectation level.

[0426] "Operation status" refers to current operational information related to operations within the organization, including task progress, resource load, abnormal situations, and process delays.

[0427] "Generative AI models" refer to AI models that can automatically generate natural language responses or structured results based on input prompts, including but not limited to deep learning-based text generation models.

[0428] "Prompt statements" refer to natural language instructions generated by the server based on user information, emotional state, and organizational data, used to instruct generative artificial intelligence models to perform specific reasoning or generation tasks.

[0429] "Recommended candidates" refers to organizational components that are determined by generative artificial intelligence models or systems to be suitable as interviewees or task support recipients.

[0430] "Suggested content" refers to textual suggestions, instructions, or guidance information that generative artificial intelligence models output based on prompts and that are related to user experience, goals, or tasks.

[0431] "Response" refers to the output generated and returned by the generative artificial intelligence model based on the prompt statement. This output may include recommended objects, suggested content, or other structured information.

[0432] "Behavioral schedule information" refers to time-schedule data associated with the components of an organization, including information recorded by schedule management services or databases such as future meetings, work assignments, and free time periods.

[0433] "Free time" refers to the time interval within the scope of the user's expected schedule and the recommended object's behavior schedule information, where neither party is occupied by other arrangements and can be used to create a new interview or provide task support.

[0434] "Schedule setting device" refers to the logical component in the server that performs program functions. This component is used to automatically determine free time and create or update schedule records based on the recommended objects and the user's schedule intentions.

[0435] "Schedule time period" refers to the specific time interval selected or adjusted during the scheduling process for arranging interviews or operational support events.

[0436] "Emotion analysis and processing" refers to the analysis and processing performed on user text data and / or voice data to determine the user's emotion category, intensity, and related characteristics.

[0437] "Emotion classification processing" refers to the identification and classification operation in emotion analysis and processing, which maps a user's emotional state to a predefined category or numerical range.

[0438] The “display interface” refers to the output interface provided by the system to present generative artificial intelligence model responses, recommendation results, and schedule information in a visual manner.

[0439] "Schedule management service" refers to a service system used to store, query, and update schedule information of users and organizational components, including network services that provide schedule reading and writing interfaces.

[0440] "Communication interface" refers to the software or hardware interface used by a server to exchange data with external services or terminals, including application programming interfaces (APIs), network protocol interfaces, etc.

[0441] "Schedule information update processing" refers to the process of writing, modifying, or deleting new interview or job support events into the schedule records of users and recommended individuals through the communication interface of the schedule management service.

[0442] The embodiments of this invention will be described in detail, combining hardware structure, software modules, data structure, and the internal processing of generative artificial intelligence models, to illustrate the system's composition and technical effects. The following embodiments are described primarily using servers, terminals, and users as the main components, but this invention is not limited thereto.

[0443] I. System Overall Structure The server includes at least one processor, storage device, and network communication interface. The processor can be a general-purpose central processing unit or a graphics processing unit. The storage device can be semiconductor memory, magnetic memory, or other computer-readable storage media for storing program code, parameters, and various data structures. The server connects to multiple terminals via a network, and the terminals can be smartphones, tablet computers, desktop terminals, or other electronic devices with display and input functions.

[0444] The terminal runs applications that display user information input and results. The server runs a set of program modules, including: a natural language processing module, a sentiment analysis module, a prompt generation module, a generative artificial intelligence invocation module, a recommendation calculation module, a schedule setting module, a data management module, and a display data generation module. The data management module persistently stores and retrieves user data, organizational structure data, and schedule data through a database management system, which can be a relational database management system.

[0445] II. Data Structure and Storage Method The server maintains multiple logical tables or datasets for each user in the storage device. For example, the server records the user identifier and basic attributes for each user in the user information table; the experience goal table records the experience text, future goal text, and update time; the emotion state table records the emotion category, intensity score, and generation timestamp; the organizational components table records the identifier of the person or task subject, skill vector, historical task records, behavior schedule references, etc.; and the schedule table records the event identifier, start time, end time, participating entity identifier, event type, etc.

[0446] The server stores the user's natural language input as raw text fields, and simultaneously stores the language processing results as structured features, including word segmentation sequences, part-of-speech sequences, named entity lists, and key information slots (such as expected date, time period, and meeting type). The server maintains a feature vector for each organizational component, which can consist of multiple dimensions, such as project experience, skill proficiency, historical performance, and communication style. These vectors are used for similarity calculation and ranking in a high-dimensional space.

[0447] III. Terminal-side processing and server communication The terminal displays multiple input controls, including text input boxes, a voice input button, and optional emotion annotation options. Users input experiences, future goals, and schedule intentions in natural language on the terminal. The terminal locally packages the user input into a request object, containing a user identifier, text content, optional audio data, and a timestamp. The terminal then sends this request object to the application programming interface exposed by the server via a secure protocol.

[0448] After receiving the server's response, the terminal displays a list of recommended people, suggested text output by the generative AI model, and specific schedule information generated by the scheduling module. The terminal can also display a visual representation of emotional states, such as stress levels or positivity levels presented in color or icon form.

[0449] IV. Server-side Natural Language Processing and Sentiment Analysis After receiving natural language input from the terminal, the server passes the text to the natural language processing module. This module can be implemented using a language processing framework. Within this module, the server performs word segmentation, part-of-speech tagging, syntactic analysis, and named entity recognition, and extracts key phrases related to professional experience, future goals, and scheduling intentions. For example, from the text "I have 5 years of production line maintenance experience and would like to be responsible for the construction of a new automated line in the future. I hope to have a one-on-one meeting with a senior engineer next Wednesday afternoon," the server extracts the numerical value "5 years," the domain terms "production line maintenance" and "automated line construction," and the scheduling constraints "next Wednesday afternoon" and "one-on-one meeting with a senior engineer."

[0450] The server uses a sentiment analysis module to classify the same text and optional speech features for emotion classification. This module can be based on a deep learning classifier, including several embedding layers, convolutional or recurrent layers, and a fully connected output layer. The server converts the text into word or sub-word sequence indices, inputs them into the embedding layer to obtain a sequence of word vectors, then extracts local emotional cues through multiple convolutional kernels, or captures contextual dependencies through a bidirectional recurrent network; finally, it outputs multi-dimensional emotion scores, such as stress scores, positivity scores, and anxiety scores, through a fully connected layer and a normalization function. The server stores these scores in an emotion state table and generates emotion labels, such as "high stress but positive intentions."

[0451] By performing the specific feature extraction and sentiment classification calculations on the server side, a high-dimensional feature representation that can be used for subsequent calculations can be generated inside the computer. This allows subsequent recommendation and scheduling calculations to rely on numerical vector operations instead of simple rules, thereby improving the parsing accuracy of natural language input at the algorithm level.

[0452] V. Prompt Statement Generation and Generative Artificial Intelligence Model Invocation The server uses a prompt generation module to combine structured user characteristics, emotional states, and organizational features into one or more natural language prompts. Following a predefined template structure, the prompt generation module embeds data from different sources into the text, enabling the generative AI model to perceive user background, current emotion, and candidate information in a single processing step.

[0453] For example, the server generates the following prompt: Below is a user's information: - Experience: 'I have 5 years of experience in production line maintenance and have participated in the implementation of 3 automated production lines.' - Goal: 'I hope to be responsible for the construction of new automated production lines in the future.' - Current mood: Under a lot of pressure but remaining positive.

[0454] The database contains the following candidate colleagues (example only): 1. Candidate A: 10 years of experience in automation projects, skilled at mentoring newcomers, with a gentle communication style.

[0455] 2. Candidate B: 8 years of automation experience, strong project execution ability, and a relatively direct personality.

[0456] 3. Candidate C: 5 years of maintenance experience, recently transitioned to automation project manager.

[0457] Based on the user's experience, goals, and emotional state, please recommend 1-2 colleagues who are best suited for one-on-one guidance and communication with the user, and explain the reasons for each recommendation (no more than 100 words). When generating such prompts, the server does not simply concatenate text, but dynamically selects the description method and granularity of detail based on the emotional state. For example, when the emotion analysis results indicate that the user is under a lot of stress, the server can add constraint statements related to psychological support; when the user's goal is ambitious but their current experience is insufficient, the server can require the model to consider "gradual guidance relationships" in the prompts.

[0458] The server sends the aforementioned prompt to the generative AI model via the generative AI invocation module. The generative AI model can be a multi-layered transform network based on a self-attention mechanism, its structure including a word embedding layer, positional encoding, several encoder layers, and a decoder layer. When invoking the interface, the server provides parameters such as model name, temperature coefficient, and maximum output length to control the diversity and length of the output.

[0459] Generative AI models internally perform multi-head attention computation, capturing the complex dependencies between "user experience—goal—emotion—candidate object" through adaptive weighting of different segments in the prompt. Compared to traditional rule engines, this attention structure can perform non-linear combinations in high-dimensional space, discovering matching patterns that are difficult to enumerate manually, thereby improving the accuracy of recommendations.

[0460] After the generative artificial intelligence model outputs recommendation results and reasoning text, the server performs grammar checks and security filtering on the output. Then, the recommendation calculation module parses out the recommendation object identifier and matches it with the entity identifier in the organizational composition element table, thereby establishing a mapping between natural language output and structured entities in the internal data structure.

[0461] VI. Collaboration between the scheduling module and external services The server obtains the behavioral schedule information of recommended users through the schedule setting module. The server can directly access the internal schedule table or access external schedule management services via a network interface. Internally, the schedule setting module uses data structures such as time interval trees or segment trees to efficiently query and perform intersection calculations on the busy and idle times of users and recommended users. When calculating idle time, the server takes the user's desired time period, the existing schedules of recommended users, and organizational rules (such as prohibiting late-night meetings) as constraints.

[0462] When the stress score in the mood status scale exceeds a predetermined threshold, the server can add mood weights during schedule calculations, such as prioritizing the exclusion of time periods following consecutive high-intensity meetings and favoring time zones with lower workloads. The server can then use a generative artificial intelligence model to help select the optimal time from multiple candidate time periods. For example, the server generates the following prompt: "The user's current stress score is 0.8, which is more suitable for communication during a more relaxed afternoon period."

[0463] Below are three optional time periods: 1) Wednesdays 09:00–09:30; 2) Wednesdays 14:00–14:30; 3) Wednesday 16:00–16:30.

[0464] Please recommend the time of day that is most conducive to relieving user stress and improving communication effectiveness, and provide a reason in no more than 50 words. The server converts the preferred time period output by the model into specific start and end times, and creates new event records for the user and recommended objects in the schedule. Furthermore, the server updates the calendar data of both parties through a communication interface with the schedule management service. In this way, the server completes end-to-end processing from user natural language input to recommended object determination, schedule setting, and calendar updates within the same computational chain, significantly reducing data round-trip transmission and manual operations between different systems, thereby reducing communication load and shortening overall response time.

[0465] VII. Emotion-Adaptive Prompt Statements and Technical Effects The server uses emotional state as a control variable during the prompt generation process, allowing the prompts to adaptively adjust in terms of content, tone, and detail. To this end, the server maintains multiple prompt templates in storage and defines a set of mapping rules: when stress is high, constraints on "relaxing communication" and "supportive communication" are added; when motivation is high, descriptions of "challenging tasks" and "rapid growth opportunities" are added. These rules are implemented in the program as conditional branches and parameter tables, rather than being determined manually on an ad-hoc basis.

[0466] This emotion-driven prompt adjustment enables generative AI models to output more appropriate suggestions for different user states within a unified architecture. Since the model has already learned language styles and strategies in different contexts during training, the server, through fine-tuning the prompts, essentially selects different generation trajectories within a high-dimensional parameter space, thus achieving "logical-level" adaptation without changing the model weights. This approach avoids the computational cost of frequent model retraining and improves overall processing speed while maintaining output quality.

[0467] VIII. Training and Updating Generative Artificial Intelligence Models In one embodiment of the present invention, the server can pre-train a generative artificial intelligence model with fine-tuning. The server collects a large number of training samples containing "user background—emotional state—candidate object information—recommendation and reasons," with each sample consisting of an input prompt and a target output text. The server uses a cross-entropy loss function to measure the difference between the model output and the target text, and updates the model parameters through a backpropagation algorithm. During training, the server can employ learning rate decay, gradient pruning, and data augmentation strategies. Data augmentation can generate equivalent training samples through methods such as synonym substitution and sentence transformation to improve the model's robustness to diverse inputs.

[0468] The model trained in this way, once deployed to a production environment, has its internal multi-layered attention weights encoding rich recommendation patterns. When the server dynamically generates prompts and invokes the model, the model internally performs highly optimized matrix multiplication and nonlinear transformation operations, which are highly efficient on hardware-accelerated architectures. Therefore, this invention improves the server's efficiency in handling complex inputs and its output stability during actual operation by combining prompt construction with model training strategies.

[0469] IX. Multiple Implementation Methods and Alternative Solutions In one implementation, the server can employ a unified generative AI model to simultaneously handle person recommendations and text suggestion generation. In another implementation, the task can be split into two models: a smaller structured decision model responsible for ranking the recommended objects, and a larger generative model responsible for generating natural language explanation text. The server can dynamically select the calling path based on system load and real-time requirements. For example, in high-concurrency scenarios, the server prioritizes calling the lightweight ranking model and generating brief explanations, thereby reducing the computational cost per request.

[0470] The server can also choose different emotion parsing architectures in different implementations, such as using a text-based emotion classification network or a multimodal network that integrates speech acoustic features and text semantic features. After concatenating the features of the multimodal input, the server feeds them into the classifier to make the emotional state more accurately reflect the user's true situation, thereby bringing higher accuracy in the stages of prompt adjustment and schedule selection.

[0471] 10. Explanation of Technical Effects and Causal Relationship Through the specific data structures, algorithmic processes, and module collaboration described above, the server achieves the following technical effects: The server integrates natural language parsing, sentiment analysis, prompt generation, generative AI invocation, and schedule calculation into a single data stream, allowing intermediate results to be transmitted in memory in the form of structured vectors and time intervals. This reduces serialization and deserialization operations between different subsystems, lowers communication overhead, and shortens processing links.

[0472] By constructing prompts for the generative AI model that include user characteristics, emotional state, and candidate object characteristics, the server enables the model's attention mechanism to complete multi-factor comprehensive reasoning in a single forward propagation. Compared with traditional multi-round rule filtering and manual screening, this reduces the number of iterations and improves the overall recommendation accuracy and response speed.

[0473] The server introduces data structures such as time interval trees in the scheduling process, which reduces the complexity of idle time calculation from linear to logarithmic or lower. This significantly improves the efficiency of time intersection calculation, especially when dealing with a large number of schedule records.

[0474] By adjusting the prompts and time slot selection strategies based on emotion-driven messages, the server effectively reduces appointments that do not match the user's psychological state, lowers the frequency of subsequent changes and cancellations, and thus indirectly reduces the total number of schedule update requests and notification messages, thereby reducing the overall communication load of the system.

[0475] In summary, this invention does not merely automate business processes, but rather achieves joint modeling and efficient computation of natural language, emotional states, and schedule information by introducing specific data structures, deep learning architectures, and unconventional rule combinations within the server. This results in multiple technical effects at the computer technology level, including improved processing accuracy, faster response times, optimized resource scheduling, and reduced communication load.

[0476] use Figure 14 The processing procedure is explained.

[0477] Step 1: Users input natural language information on the terminal. Users open the application interface on the terminal, enter natural language content about their own experience, future goals, and schedule intentions in the text input box, and can choose to add voice data via voice input. An example input is: "I have 5 years of experience in production line maintenance and want to be responsible for the construction of a new automated line in the future. I hope to have a chat with a senior engineer next Wednesday afternoon." Input: Raw natural language text (optional voice data), user identifier.

[0478] Output: A locally generated request data object on the terminal, which includes a user identifier, text content, optional audio binary data, and a timestamp.

[0479] Based on the user's "submit" action, the terminal encapsulates the request data object into a network request message, ready to send it to the server.

[0480] Step 2: The terminal sends a request to the server and receives a preliminary confirmation. The terminal uses a security protocol via the network communication interface to send the request data object generated in step 1 to the application programming interface address provided by the server. The terminal attaches an authentication token to the request header so that the server can verify the user's identity.

[0481] Input: Request data object (containing user text, audio, user ID, and authentication information).

[0482] Output: The network message sent to the server, and the sending status (success / failure) on the terminal.

[0483] After receiving the request, the server returns a simple confirmation response to the terminal, which includes the request number and the receipt status. The terminal can temporarily display a "Processing" status message on the interface.

[0484] Step 3: The server performs request parsing and data verification. After receiving the request message from the terminal at the network layer, the server passes it to the backend processing module. The server parses the JSON or equivalent format, extracts the user identifier, natural language text, and optional voice data, and checks the integrity of the fields and the validity of the data format.

[0485] Input: A network request message received from the terminal.

[0486] Output: An internal standardized request object (containing user ID, text field, audio field, and timestamp), and the validation result (valid / invalid).

[0487] If the server detects a missing field or invalid data, it generates an error response and returns it to the terminal; if the verification passes, it proceeds to the subsequent language processing and sentiment analysis process.

[0488] Step 4: The server stores the original input in the database and creates a request record. The server uses the data management module to connect to the database, writing information such as the user's original text, optional audio reference path, and request time into the request record table, and assigning a unique request ID to each record. Simultaneously, the server updates the user information table with the most recent interaction time and related status.

[0489] Input: Standardized request object (user ID, text, audio, timestamp).

[0490] Output: Request record entries (including request IDs) stored in the database, and the mapping between request IDs and user IDs held in memory.

[0491] The server persists the data by executing insert statements, establishing foreign key relationships between the subsequent parsing results and recommendation results.

[0492] Step 5: The server performs natural language processing and extracts structured features. The server invokes the natural language processing module to perform word segmentation, part-of-speech tagging, named entity recognition, and dependency parsing on the user's text. Based on the analysis results, the server extracts keywords, numerical information (such as years), date and time expressions, and role descriptions related to "experience," "future goals," and "schedule intentions" from the text.

[0493] Input: User's natural language text (string) and language processing configuration parameters.

[0494] Output: Structured language feature objects, including word segmentation sequences, entity lists, experience slots (e.g., "5 years of production line maintenance"), target slots (e.g., "person in charge of automation line construction"), and schedule slots (e.g., "1-on-1 meeting with a senior engineer next Wednesday afternoon").

[0495] The server stores these features in the parsing results table, forming the intermediate data required for subsequent calculations.

[0496] Step 6: The server performs emotion analysis on user input and generates emotion states. The server performs emotion classification processing on the user's text and optional speech features in the emotion parsing module. The server converts the text into a vector sequence and inputs it into a pre-trained emotion classification neural network model, and extracts acoustic features (such as pitch and energy changes) from the audio as additional input.

[0497] Input: User text content, audio feature data (if available), and emotion classification model parameters.

[0498] Output: Emotional state object, including emotion category (e.g., "highly stressed but positive"), multiple emotion scores (stress score, positivity score, anxiety score, etc.) and timestamp.

[0499] The server writes the emotional state into an emotional state table and binds it to the current request ID in memory, providing control variables for subsequent prompt generation and schedule selection.

[0500] Step 7: Server retrieval organizational components and candidate object characteristics Based on the target slots and experience slots extracted in step 5, the server retrieves entities with relevant skills and experience from the organizational composition element table, such as personnel or operational entities with experience in automation projects. The server reads the corresponding feature vectors (skills, years of experience, communication style, etc.) and basic information.

[0501] Input: User structured characteristics (domain information, target roles, etc.) and database content of organizational components.

[0502] Output: A list of candidate objects, including the identifier, feature vector, and brief description of each candidate entity, which will be presented in subsequent prompts.

[0503] The server can limit the candidate list to a certain range based on preliminary similarity matching or rule filtering, in order to control the input length of the generative artificial intelligence model.

[0504] Step 8: The server generates prompts for invoking the generative artificial intelligence model. The server invokes the prompt generation module, embedding user experience text, target text, emotional state description, and candidate object summary into a predefined text template to form a structured and detailed prompt. The server adjusts the tone, constraints, and level of detail of the prompt based on the emotional state.

[0505] Input: User's original text, structured features (experience, goals, schedule intentions), emotion state object, and candidate object list.

[0506] Output: One or more natural language prompts, which are submitted to the generative AI model to perform recommendation and suggestion generation tasks.

[0507] For example, the prompt message generated by the server can be: Below is a user's information: - Experience: 'I have 5 years of experience in production line maintenance and have participated in the implementation of 3 automated production lines.' - Goal: 'I hope to be responsible for the construction of new automated production lines in the future.' - Current mood: Under a lot of pressure but remaining positive.

[0508] The database contains the following candidate colleagues: 1. Candidate A: 10 years of experience in automation projects, skilled at mentoring newcomers, with a gentle communication style.

[0509] 2. Candidate B: 8 years of automation experience, strong project execution ability, and a relatively direct personality.

[0510] 3. Candidate C: 5 years of maintenance experience, recently transitioned to automation project manager.

[0511] Based on the user's experience, goals, and emotional state, please recommend 1-2 colleagues who are best suited for one-on-one guidance and communication with the user, and explain the reasons for each recommendation (no more than 100 words). Step 9: The server invokes a generative artificial intelligence model and obtains a response. The server sends the prompt as input to the generative AI model interface through the generative AI invocation module, and sets the generation parameters (such as temperature coefficient and maximum output length). Internally, the generative AI model performs forward computation based on a multi-head attention mechanism to generate a natural language response containing a list of recommendations and reasons.

[0512] Input: Prompt text, model configuration parameters.

[0513] Output: The response text returned by the generative AI model, which includes the name or description of the recommended object, the reason for the recommendation, and possible additional suggestions.

[0514] After receiving the response, the server performs security checks and basic format standardization on the returned text to prepare for parsing and mapping.

[0515] Step 10: The server parses the model's response and maps it to structured recommendation results. In the recommendation calculation module, the server parses the response text of the generative artificial intelligence model, identifies the name or identifier of the recommended object, and matches it with the corresponding entity records in the organizational composition element table. The server uses string matching, similarity calculation, or pre-embedded tags to map natural language descriptions to specific entity IDs.

[0516] Inputs: Model response text, candidate object list, and organizational composition database.

[0517] Output: A structured recommendation result object, containing a set of recommended entity IDs, the text of each recommendation reason, and recommendation level information.

[0518] The server writes the recommendation result into the recommendation result table and associates it with the current request ID, providing a data source for schedule setting and subsequent display.

[0519] Step 11: The server retrieves the user's and recommended users' schedule information and calculates their free time. The server calls the scheduling module to read the user's and recommended users' schedules near the target date from the calendar or an external schedule management service. The server organizes the busy / slack status of each entity according to the time interval structure and calculates the intersection time period based on the user's schedule intention (such as "next Wednesday afternoon").

[0520] Input: User's schedule intention (structured time slots), list of recommended object IDs, schedule data source (internal table or list of events returned by external service).

[0521] Output: A list of shared free time slots for both the user and the recommended users, with each time slot including start time, end time, and a schedule tag.

[0522] The server can filter out unsuitable time periods based on internal rules, such as late at night or outside of working hours.

[0523] Step 12: The server selects the optimal appointment time based on the user's emotional state and available time. After obtaining multiple shared idle time periods, the server combines these time periods with emotional states for optimal selection. The server can execute a scoring-based ranking algorithm locally, or it can further utilize generative artificial intelligence models to assist in decision-making.

[0524] Input: List of available time slots, mood state object, and optional additional business rules.

[0525] Output: One or a few preferred time periods, along with the corresponding explanations.

[0526] For example, a server can send the following prompt to a generative artificial intelligence model: "The user's current stress score is 0.8, which is more suitable for communication during a more relaxed afternoon period."

[0527] Below are three optional time periods: 1) Wednesdays 09:00–09:30; 2) Wednesdays 14:00–14:30; 3) Wednesday 16:00–16:30.

[0528] Please recommend the time of day that is most conducive to relieving user stress and improving communication effectiveness, and provide a reason in no more than 50 words. The server determines the final booking time based on the recommended time slot returned by the model and internal constraints.

[0529] Step 13: The server creates schedule events and updates schedule information. After the server selects the final appointment time and the recommended participants, it creates new event records for both the user and the recommended participants in the calendar, including start time, end time, event type, title, and description. If the server is integrated with an external calendar management service, it performs event insertion or update operations through its communication interface to ensure that the personal calendars of the user and the recommended participants are synchronized.

[0530] Input: Final appointment time slot, user ID, referral ID, and event description.

[0531] Output: The ID of the newly created or updated schedule record, and the status of the operation result from the external schedule service.

[0532] The server internally records the schedule creation results and generates a unified schedule information object for subsequent notifications and displays.

[0533] Step 14: The server generates feedback text and display data and sends them to the terminal. Based on recommendations, scheduling results, and emotional state, the server generates a summary response. The server can then leverage generative AI models to generate brief, user-facing feedback, such as affirmations of their experiences and goals, and suggestions for next steps.

[0534] Inputs: Recommended results object, schedule information object, emotional state, and optional feedback strategies.

[0535] Output: Uniform response data, including: a list of recommended individuals or tasks, reasons for recommendation, confirmed appointment time and location / method, and emotional feedback text.

[0536] The server returns the response to the terminal via the network. The terminal displays this information on the interface as a list and text description, allowing the user to view the planned meeting time and participants, and confirm or modify them as needed.

[0537] Step 15: Users confirm or adjust appointments on the terminal and trigger subsequent interactions. After viewing the recommended attendees and schedule information returned by the server on the terminal, users can click "Confirm Reservation" to finalize the arrangement, or modify the time or add notes through the terminal interface. The terminal then resends the user's confirmation or adjustment to the server.

[0538] Input: The user's confirmation or adjustment instructions for the reservation result.

[0539] Output: Status update request messages sent from the terminal to the server, and local interface updates on the terminal.

[0540] After receiving the update request, the server modifies the schedule records and related statuses, and calls the schedule setting module again to synchronize with the external schedule service as needed, thereby maintaining data consistency and a closed loop of technical processing throughout the process.

[0541] 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.

[0542] 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.

[0543] 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.

[0544] 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.

[0545] 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.

[0546] Second Implementation Method Figure 3 An example of the configuration of the data processing system 210 according to the second embodiment is shown.

[0547] 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.

[0548] 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).

[0549] 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.

[0550] 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.

[0551] 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).

[0552] 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.

[0553] 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.

[0554] 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.

[0555] 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).

[0556] 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.

[0557] 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".

[0558] 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.

[0559] 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.

[0560] 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.

[0561] 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.

[0562] 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.

[0563] 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.

[0564] 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.

[0565] 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.

[0566] 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.

[0567] Third Implementation Method Figure 5 An example of the configuration of the data processing system 310 according to the third embodiment is shown.

[0568] 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.

[0569] 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).

[0570] 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.

[0571] 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.

[0572] 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).

[0573] 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.

[0574] 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.

[0575] 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.

[0576] 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.

[0577] 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.

[0578] 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".

[0579] 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.

[0580] 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.

[0581] 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.

[0582] 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.

[0583] 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.

[0584] 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.

[0585] 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.

[0586] 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.

[0587] 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.

[0588] Fourth Implementation Method Figure 7 An example of the configuration of the data processing system 410 according to the fourth embodiment is shown.

[0589] 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.

[0590] 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).

[0591] 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.

[0592] 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.

[0593] 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).

[0594] 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.

[0595] 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.

[0596] 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.

[0597] 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.

[0598] 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.

[0599] 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.

[0600] 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".

[0601] 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.

[0602] 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.

[0603] 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.

[0604] 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.

[0605] 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.

[0606] 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.

[0607] 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.

[0608] 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.

[0609] 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.

[0610] 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.

[0611] 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.

[0612] 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.

[0613] 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).

[0614] 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.

[0615] 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."

[0616] 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.

[0617] 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).

[0618] 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.

[0619] 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.

[0620] 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.

[0621] 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.

[0622] 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.

[0623] 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.

[0624] 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.

[0625] 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.

[0626] 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.

[0627] 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.

[0628] In addition, the following notes are provided in response to the above explanation.

[0629] Example 1 (Note 1) An information processing system, characterized in that it comprises: A unit for generating an input / output interface on a display device for acquiring user career experience information and future goal information, and a unit for receiving the career experience information and future goal information as text information through the input / output interface; Units used to convert the text information into a data structure, verify the existence, length and format of the text information according to a predetermined data format, and store the verified text information as structured data. This unit is used to perform natural language processing on the structured data, including word segmentation, word extraction, and attribute classification, to extract current capability information and target direction information from the structured data, and to generate a prompt statement as input to a generative artificial intelligence model by inserting the text information and the extraction results into a predefined template based on the current capability information and the target direction information. A unit for sending the generated prompt statement to the generative artificial intelligence model configured externally or internally via a program interface according to a communication protocol, and for receiving response data obtained from the generative artificial intelligence model; A unit for extracting professional competence information, learning project information and action plan information from the response data, organizing and editing the extracted information by project into display data for visualization, and displaying the display data in partitions on the input / output interface of the display device. A unit for recording the text information, the prompt statement, and the response data as historical information in a storage device, and updating and maintaining the template for generating the prompt statement or the parameters of natural language processing based on the historical information.

[0630] (Note 2) The information processing system according to Appendix 1 is characterized in that, The unit for performing natural language processing is configured to extract category information such as occupation category, skill category, and years of work experience from the structured data, select a prompt statement template from a variety of prompt statement templates based on the category information, and generate prompt statements according to the selected prompt statement template to specify the processing content to be performed in the generative artificial intelligence model in paragraph and enumeration structures.

[0631] (Note 3) The information processing system according to Appendix 1 is characterized in that, The unit for editing and displaying data is configured to parse the response data obtained from the generative artificial intelligence model, generate differential information representing the difference between the user's current ability information and the recommended ability information in the response data, and learning order information arranged along the time axis, and display the differential information and the learning order information on the display device in a list or hierarchical form.

[0632] Application Example 1 (Note 1) An information processing system, characterized in that it comprises: A processing unit that functions as a processing unit in an information processing device; The processing unit is configured to display a means on a communication terminal for allowing the user to input experience information and target information as text information; The processing unit is configured to convert the text information obtained from the communication terminal into a data format for storage and store it in a recording medium in a storage device on a user-by-user basis. The processing unit is configured to, based on the stored experience information and target information, insert the experience information and target information into a predefined template to form a prompt statement that can be used as input for a generative artificial intelligence model. The processing unit is configured to send the constructed prompt statement to the generative artificial intelligence model via a communication network, and to obtain the generated text data from the generative artificial intelligence model. The processing unit is configured to perform filtering processing on the generated text data based on keyword determination and paragraph segmentation, and to extract the more useful parts from the generated text data to generate structured information data according to evaluation rules associated with user interest attributes. The processing unit is configured to convert the structured information data into display data for visualization on the display unit of the communication terminal, and to distribute it to the communication terminal.

[0633] (Note 2) The information processing system according to Appendix 1 is characterized in that, The processing unit is configured to, when generating the prompt statement, in addition to the stored experience information and target information, also parse the feedback information obtained from the user and the generation history information of the generative artificial intelligence model, and adjust the prompt statement for different users one by one by updating the generation conditions in the template.

[0634] (Note 3) The information processing system according to Appendix 1 is characterized in that, The processing unit is configured to, when performing the filtering process on the generated text data, divide the generated text data into multiple information units, perform a scoring calculation on each information unit based on the user's professional field information and target time information, and select or discard the information units according to the result of the scoring calculation to form the structured information data.

[0635] Example 2 (Note 1) An information processing system, characterized in that it comprises: Means for providing an input / output interface for user experience information and future target information by a processing unit on an information processing device; A means for parsing natural language text and generating feature quantities representing user needs based on the user experience information, the future goal information, and the prompts input by the user. Means for generating search conditions and obtaining member candidate data from an information storage device that records members within an organization and their associated outputs based on the aforementioned feature quantity; A means for preprocessing the member candidate data and the user's feature values ​​to generate input data for learning or inference, including numerical vectors. A means for using the input data to calculate the fit of each member candidate through a machine learning model, and to generate a member candidate list for the user based on the fit. Means for generating prompt text for input into a generative artificial intelligence model based on the user's prompt statement and the member candidate list, and sending the prompt text to the generative artificial intelligence model to obtain response text containing recommendation reasons or explanatory information; Means for visually displaying recommendation results containing the member candidate list and the response text through the input / output interface; A means for obtaining the schedule information of the members selected by the user and the user, calculating their free time using a schedule management program interface and automatically registering the meeting schedule.

[0636] (Note 2) According to the information processing system described in Appendix 1, the processing unit is configured to: construct the machine learning model as a neural network model that takes member attribute information, outcome information, user experience information, and future goal information as inputs and outputs the fitness of each member; and use natural language processing technology to convert the text data input to the neural network model into numerical vectors.

[0637] (Note 3) According to the information processing system described in Appendix 1, the processing unit is configured to: generate, in association with the member candidate list, the response text obtained from the generative artificial intelligence model as explanatory information corresponding to each recommended member, and generate a user interface containing the explanatory information for display, so as to support the user in performing member selection operations.

[0638] Application Example 2 (Note 1) An information processing system, characterized in that it comprises: A device for providing a user information input interface configured to receive natural language input from users regarding their experiences, future goals, and schedule intentions; A device for performing language processing on natural language input obtained from the user, the language processing being configured to extract candidate persons or task subjects from the constituent elements within the organization based on the user's experience and future goals, and to generate prompt statements that instruct a generative artificial intelligence model to perform the recommendation of persons or task subjects and the generation of task assignments or suggestions, together with additional information including the user's emotional state and task status. A device for sending the prompt statement to the generative artificial intelligence model and obtaining a response from the generative artificial intelligence model, including recommended objects or suggested content; A schedule setting device for acquiring behavioral schedule information of organizational components corresponding to the recommended object included in the response, determining the free time of both parties based on the user's expected schedule and the behavioral schedule information, and automatically setting the interview or job support schedule during the free time. A means for inferring the user’s emotional state based on the user’s natural language input and the response from the generative artificial intelligence model, and for adjusting at least one of the content of the prompt statement and the schedule time period set by the schedule setting device based on the emotional state. A device for providing a display interface configured to visually output the response from the generative artificial intelligence model and schedule-related information set by the schedule setting device.

[0639] (Note 2) The information processing system according to Appendix 1 is characterized in that, The device for estimating the emotional state is configured to perform emotion analysis processing on the user's text data and voice data, and to modify the tone, level of detail, or recommendation criteria of the prompts sent to the generative artificial intelligence model based on the result of the emotion analysis processing.

[0640] (Note 3) The information processing system according to Appendix 1 is characterized in that, The scheduling device is configured to select candidate individuals or candidate task subjects from the candidate set obtained from the generative artificial intelligence model that are related to the recommended object, a behavioral schedule information database maintained within the organization, the user's expected conditions, and the emotional state, and update the schedule information of the user and the candidate individuals or candidate task subjects through the communication interface of the schedule management service according to the selection result.

Claims

1. An information processing system, characterized in that, include: processor; The processor is configured to provide the user with an interface for inputting user experience and the user's future goals; The input information received by the interface is parsed, and prompts are generated to instruct the generative artificial intelligence model to perform specific processing. The generated prompt information is sent to the generative artificial intelligence model, and a response is received from the generative artificial intelligence model.

2. The information processing system according to claim 1, characterized in that, The processor is configured to: parse the user's input data using natural language processing technology, and generate prompts based on the parsing results to instruct the generative artificial intelligence model to perform optimal processing.

3. The information processing system according to claim 1, characterized in that, The processor is configured to visualize the response from the generative artificial intelligence model as information related to the user's career development.

Images