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: based on the set theme of the department, generate prompt information for setting the goals of each job level by using a generative artificial intelligence model, and generate the corresponding goals according to the prompt information; based on the input achievement content, generate prompt information for one-time judgment of the achievement degree by using a generative artificial intelligence model, and perform one-time judgment of the achievement degree according to the prompt information; for identifying the user emotion, analyze the voice and expression of the user by using an emotion analysis engine, and supplement the judgment of the achievement degree according to the identified emotion.

Description

Technical Field

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

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

[0003] In organizational performance evaluation and goal management, traditional goal setting and achievement assessment rely heavily on human experience and subjective judgment, leading to the following problems: First, goal setting is often based solely on the manager's personal experience, failing to fully utilize historical performance data and departmental characteristics. This results in goals for different job levels within the same department that are not scientific or objective enough, and cannot effectively reflect differentiated requirements. Second, achievement assessment typically relies on manual scoring or simple quantitative comparisons, lacking semantic understanding and comprehensive analysis of the achieved content, making it difficult to provide a reasonable judgment on complex work results. Third, in actual evaluations, users' emotional states (e.g., tension, frustration, or confidence) are often overlooked. The evaluation results fail to consider the user's emotional information during the performance review process, resulting in a single evaluation dimension that cannot comprehensively reflect the user's actual performance. Furthermore, existing systems often display evaluation results in text or simple numerical lists, which is not conducive to individuals quickly and intuitively understanding their achievement status, thus affecting subsequent improvement and growth planning. Therefore, a system is needed that combines generative artificial intelligence models and emotion analysis technology to automatically generate goals for each job level, intelligently judge the achieved content, supplement the evaluation with user emotions, and display the results in intuitive charts to solve the above problems. Summary of the Invention

[0004] To address the aforementioned issues, this invention provides an information processing system comprising a processor configured to: generate prompts for setting goals at various job levels based on a departmental theme using a generative artificial intelligence model, and generate corresponding goals based on these prompts, thereby fully utilizing historical data and departmental characteristics during the goal-setting phase to achieve automated and differentiated goal setting for different job levels; generate prompts for determining the degree of achievement based on the input achieved content using a generative artificial intelligence model, and perform a first-order determination of achievement based on these prompts, so as to perform semantic understanding and quantitative analysis on the achieved content submitted by the user in natural language or structured form, and output an objective first-order determination result; and, to identify user emotions, analyze the user's voice and facial expressions using an emotion analysis engine, and supplement the degree of achievement determination based on the identified emotions, thereby introducing an emotion dimension into the evaluation process, comprehensively considering the user's state, and improving the comprehensiveness and rationality of the evaluation results. Furthermore, the processor can generate prompts for a second assessment of achievement based on the initial assessment result using a generative artificial intelligence model, and perform the second assessment according to these prompts, thus verifying and deeply evaluating the initial assessment. Additionally, the processor can display the achievement assessment result in graphical or chart form using an information display device, allowing individuals to easily understand the result and thus enabling them to promptly grasp their performance level and develop targeted improvement plans. Through the above structure and processing flow, this invention comprehensively introduces generative artificial intelligence models and emotion analysis technology into all aspects of goal setting, achievement assessment, and result display, effectively solving the technical problems of unscientific goal setting, single evaluation dimensions, and unintuitive results in traditional assessment systems.

[0005] "System" refers to an overall device or platform consisting of hardware and / or software, used to perform a series of processing steps such as goal setting, achievement determination, emotion recognition and result display, which can be deployed on a local server, cloud environment or distributed environment.

[0006] A "processor" is a computing unit that can execute program instructions, process input data and output results. It can be a physical CPU, GPU, application-specific integrated circuit, microcontroller, or a processing module consisting of multiple processing cores, or a logical processing unit implemented through virtualization.

[0007] "Department-specific themes" refer to the business priorities or work directions that an organization or department predetermines within a specific assessment cycle. These themes guide the goal setting of employees at all levels within the department, such as "increasing quarterly sales" or "improving product stability."

[0008] "Job grade goals" refer to quantitative or qualitative work performance indicators set for each job grade based on the responsibilities and ability requirements of employees at different job grades. These indicators are used to measure the performance of employees at that job grade within a specific period.

[0009] "Generative AI models" refer to AI models based on technologies such as deep learning that can automatically generate text, numerical values, or other forms of output based on input prompts. For example, large language models are used to generate target content, determine prompts, or evaluate text.

[0010] "Prompt information" refers to instructions, questions, contextual descriptions, or constraints input into a generative artificial intelligence model to guide it in producing outputs of the expected type and style.

[0011] "Input Achievements" refers to the work results submitted by the user at the end of the assessment period or at a specified time, in response to the established goals. These results may include natural language text descriptions, structured data, or numerical indicators.

[0012] "Initial assessment of achievement" refers to the initial evaluation process conducted by the system using a generative artificial intelligence model based on the input content and corresponding goals, to obtain a preliminary assessment result of the goal completion status, which may be expressed in the form of a score, percentage, or grade.

[0013] An “emotion analysis engine” refers to a software module or hardware device used to analyze a user’s voice, facial expressions, or video signals to identify the user’s current emotional state (such as positive, negative, neutral, tense, confident, etc.), which can be achieved based on technologies such as voice emotion recognition and facial expression recognition.

[0014] "User's voice and facial expressions" refers to the voice signals and facial expression images or video data generated during the process of users giving presentations, providing feedback, or interacting with the system, which are used to reflect the user's emotional state at that time.

[0015] "Supplementing the achievement determination based on the identified emotions" means that after completing one achievement determination, the system uses the emotion analysis results to revise the determination, add explanations or risk warnings, thereby comprehensively considering the user's emotional factors in the evaluation.

[0016] "Secondary assessment of achievement" refers to the evaluation process in which the system uses a generative artificial intelligence model to conduct in-depth analysis or review based on the initial assessment result, in order to form a more comprehensive, refined or hierarchical final assessment result.

[0017] "The result of achievement assessment" refers to the final evaluation information on the achievement of the target obtained through one and / or two assessments, which may include scores, percentages, grades, written descriptions, and suggestions for improvement.

[0018] "Information display device" refers to an output device used to present information such as text, images, and charts to users, including but not limited to monitors, LCD screens, touch screens, mobile terminal screens, or other visual display media.

[0019] "Displaying in the form of graphs or charts" refers to presenting the achievement determination results in a visual format using visualization methods such as bar charts, line charts, pie charts, radar charts, and dashboards, so that users can intuitively understand the relationships and differences between the data. Attached Figure Description

[0020] Figure 1 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the first embodiment.

[0021] 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.

[0022] Figure 3 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the second embodiment.

[0023] 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.

[0024] Figure 5 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the third embodiment.

[0025] 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.

[0026] Figure 7 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the fourth embodiment.

[0027] 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.

[0028] Figure 9 This represents an emotion map that maps multiple emotions.

[0029] Figure 10 This represents an emotion map that maps multiple emotions.

[0030] Figure 11This is a sequence diagram illustrating the processing flow of the data processing system of the first embodiment.

[0031] Figure 12 This is a sequence diagram illustrating the processing flow of the data processing system in Application Example 1.

[0032] Figure 13 This is a sequence diagram illustrating the processing flow of the data processing system of the second embodiment.

[0033] Figure 14 This is a sequence diagram illustrating the processing flow of the data processing system in Application Example 2. Detailed Implementation

[0034] Hereinafter, an example of an implementation of the system according to the present disclosure will be described with reference to the accompanying drawings.

[0035] First, let me explain the terminology used in the following instructions.

[0036] 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.

[0037] 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.

[0038] 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.

[0039] 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.

[0040] 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.

[0041] First Implementation Method Figure 1 An example of the configuration of the data processing system 10 according to the first embodiment is shown.

[0042] 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.

[0043] 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).

[0044] 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.

[0045] 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.

[0046] 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.

[0047] 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.

[0048] Figure 2 The diagram shows an example of the main functions of the data processing device 12 and the smart device 14.

[0049] 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.

[0050] 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).

[0051] 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.

[0052] 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.

[0053] 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."

[0054] In organizational goal management and performance evaluation, the reliance on human experience to set goals and assess achievement levels presents several technical challenges: First, traditional information processing systems often simply store user-inputted goal conditions or perform rule matching, lacking in-depth modeling of historical data. This makes it difficult to automatically generate structured and quantitative goal information based on different levels and themes, resulting in a lack of objectivity and consistency in goal setting. Second, even when existing systems incorporate machine learning or generative artificial intelligence models, they often rely on single-call natural language generation, lacking a complete computational process that integrates structured goal parameter calculation, automatic construction of prompt statements, and the generation model. This makes it difficult to stably output executable natural language goal text under business constraints. Third, achievement assessment in many systems still relies primarily on static rules or single numerical thresholds. This approach fails to fully utilize generative artificial intelligence models for flexible text-level assessments or effectively integrate emotional information from user voice and facial expressions. Consequently, it cannot dynamically correct assessment results at the information processing level, hindering the adaptation of assessment results to actual achievement levels. Furthermore, user feedback on the system's generated targets and judgment results is often only used as display information, rather than being systematically used as training data and a basis for model updates. This results in insufficient adaptability of the system to changes in the environment and organizational behavior patterns, and a lack of a closed-loop self-optimizing computing architecture.

[0055] In other words, existing technologies lack an end-to-end framework for goal setting and achievement assessment, encompassing data reception, standardized storage, historical data modeling, structured target parameter calculation, automatic generation of prompts, generative AI model invocation, sentiment analysis and correction, visualization, and feedback-driven model updates. How can we improve the data processing flow and model invocation methods on the server side to achieve this? (1) The input topic information and hierarchical information can be standardized into an internal representation suitable for machine learning processing and efficiently associated with historical records; (2) Calculate structured target parameters automatically based on historical target and achievement data, and make technical corrections through constraints; (3) Automatically construct high-quality prompt statements to drive the generative artificial intelligence model to output natural language target text that conforms to business constraints; (4) The first judgment of the generative artificial intelligence model is fused with the sentiment analysis result at the computer level to form a more accurate achievement judgment; (5) Utilize user feedback and judgment results to periodically update the model and prompt generation rules in a data-driven manner, forming an adaptive computing system. This has become a computer technology problem that needs to be solved in this field.

[0056] 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.

[0057] In this invention, the server includes: a device for receiving topic information and hierarchical information as conditions for setting organizational goals, receiving the topic information and hierarchical information as input information from a terminal, validating the content of the input information, converting the input information into a standardized internal representation, and storing it in association with a storage device; a device for calculating target parameters by using a trained model trained with a machine learning algorithm based on historical information about past goals and achievement levels stored in the storage device, and applying threshold conditions and constraints to the target parameters to generate structured target information including period information and achievement indicators; and a device for automatically generating prompt statements for input to a generative artificial intelligence model based on the structured target information, the topic information, and the hierarchical information, inputting the prompt statements into the generative artificial intelligence model to generate target text in natural language, and performing inappropriate expression checks on the generated target text. The apparatus includes: a device for processing and determining achievement; a device for outputting the target text to a display device for presentation on a terminal, obtaining evaluation information or feedback information as a correction request from the terminal, storing the feedback information in association with historical information, and accumulating it as learning data for updating the trained model and prompt statement generation processing; a device for generating a prompt statement for inputting into a generative artificial intelligence model based on actual information as input to the achievement content, obtaining a first-order achievement result using the generative artificial intelligence model based on the first-order prompt statement, inferring the emotional state from the user's voice and facial expression information using an emotion analysis algorithm, correcting the first-order achievement result based on the emotional state, thereby generating an achievement determination result; and a device for generating display data that displays the achievement determination result and the target information in a graphical or chart-like manner on a visualization information display device, and presenting it in a form that is easy for the user to understand at a glance. This allows for the formation of an end-to-end computing process on the server side, encompassing data acquisition, internal representation standardization, historical data modeling, structured target parameter calculation, automatic construction of prompt statements, generative artificial intelligence model invocation, emotional information fusion and judgment, and result visualization and feedback-driven model updates. This improves the automation, objectivity, consistency, and adaptability of the goal setting and achievement judgment process, as well as the system's ability to adapt to environmental changes, at the computer technology level.

[0058] "Thematic information" refers to textual information used to express the business direction, work priorities, or management policies that an organization hopes to achieve within a specific period, including but not limited to descriptions of goals such as performance improvement, cost control, quality improvement, and customer satisfaction enhancement.

[0059] "Hierarchical information" refers to the classification information used to indicate the position of personnel within an organization in the organizational structure, including but not limited to job level, position level, management level, or job category, used to distinguish different scopes of responsibility and difficulty of objectives.

[0060] "Terminal" refers to a human-computer interaction device used by users to input data and view results, including but not limited to computer equipment, mobile communication equipment, or other electronic devices with display and input functions.

[0061] "Input information" refers to the data content sent by the terminal and received by the server for setting goals or determining the degree of achievement, including topic information, hierarchical information, achieved content, and user feedback.

[0062] "Standardized internal representation" refers to the internal data structure or feature representation that is suitable for computer programs and machine learning models to process after the original input information has undergone format conversion, encoding, cleaning and standardization.

[0063] "Storage device" refers to a data storage component used to persistently store input information, historical information, model parameters and intermediate processing results, including but not limited to database systems, file storage systems or other non-volatile storage media.

[0064] "Historical information" refers to a set of data accumulated and stored in a storage device over a period of time, which is related to past goal setting, actual achievement, and user feedback, and is used to train and update models.

[0065] "Machine learning algorithms" refer to computational methods that learn patterns from historical data through training and use them to predict or infer new data output results, including but not limited to regression, classification, clustering, or deep learning algorithms.

[0066] A "trained model" refers to a computational model with fixed or updatable parameters obtained by training it using machine learning algorithms based on historical information. It is used to output target parameters or judgment results based on input features.

[0067] "Target parameters" refer to key quantitative or classification indicators used to describe the target, which are calculated by the trained model based on topic information and hierarchical information, including but not limited to improvement magnitude, time period, difficulty level, or weight coefficient.

[0068] "Threshold conditions" refer to boundary conditions or restriction rules used to constrain the numerical range of target parameters, preventing the generation of unreasonable parameter values ​​that are too high or too low.

[0069] "Constraints" refer to a set of logical restrictions or conditions imposed on target parameters or output results in order to meet business rules, policy requirements, or system settings, in addition to threshold conditions.

[0070] "Structured target information" refers to target information represented by a predetermined data structure, obtained by organizing and encoding target parameters, and includes at least period information, achievement indicators, and related auxiliary attributes.

[0071] "Period information" refers to data content used to indicate the applicable time range of a target, including but not limited to time intervals such as days, weeks, months, quarters, or years.

[0072] "Achievement indicators" refer to quantifiable evaluation items used to measure the completion of objectives, including but not limited to sales, number of units, ratios, ratings, or other calculable performance indicators.

[0073] "Generative AI models" refer to AI models that can automatically generate text output or other content based on input text or other conditions, including but not limited to deep learning-based text generation models.

[0074] "Prompt statements" refer to input text constructed to drive generative artificial intelligence models to output content that meets expectations, including descriptions of the task, background conditions, and output requirements.

[0075] "Target text" refers to a target description statement expressed in natural language, generated by a generative artificial intelligence model based on prompts.

[0076] "Inappropriate expression detection and judgment processing" refers to the process of analyzing the generated target text to detect and judge whether it contains sensitive content, illegal expressions, obvious logical errors, or content that does not conform to predetermined rules.

[0077] "Evaluation information" refers to the subjective evaluation data given by users to the target text or judgment results generated by the system, including but not limited to satisfaction ratings, options, or short labels.

[0078] "Feedback information" refers to suggestions for correction, supplementary explanations, or objections raised by users regarding the generated target text, which reflects users' actual feelings about the results and their business needs.

[0079] "Achievements" refers to the actual results or performance data corresponding to the established goals, used to reflect the actual completion of the goals within a specific period.

[0080] "Actual information" refers to various types of data recorded and entered by users or systems to describe objective execution results, including numerical records, text descriptions, or file records.

[0081] "Prompt statements for initial judgment" refers to prompt statements constructed for generative artificial intelligence models to perform their first achievement judgment, which include the achievement content, target information, and judgment requirements.

[0082] "First-order judgment result" refers to the preliminary judgment result of the generative artificial intelligence model based on the first-order judgment, which is output by prompting statements, on whether the achieved content has achieved the goal. It may include text descriptions, ratings, or labels.

[0083] "Sentiment analysis algorithms" refer to algorithms used to automatically identify a user's emotional state from data such as voice, images, or text, including but not limited to sentiment recognition methods based on feature extraction and classifiers.

[0084] "Emotional state" refers to the category of a user's emotions or attitudes at a specific moment, as inferred by an emotional analysis algorithm, including but not limited to positive, neutral, negative, tense, or angry.

[0085] "Achievement determination result" refers to the systematic determination conclusion obtained by combining emotional state and relevant rules with an initial determination result, and is used to finally represent the degree of goal completion.

[0086] "Information display device" refers to an output component used to present target information and judgment results to a user in a visual manner, including but not limited to a display, electronic screen or projection device.

[0087] "Display data" refers to a formatted, visual data structure generated for presentation on an information display device in the form of graphics, charts, or other visual means.

[0088] "Prompt statements for secondary judgment" refer to prompt statements constructed after obtaining the initial judgment result and emotional state, in order to further utilize generative artificial intelligence models for secondary judgment. They typically include the initial judgment result, emotional information, and adjustment requirements.

[0089] "Secondary judgment result" refers to the further judgment result of the achievement level output by the generative artificial intelligence model based on the secondary judgment using prompt statements, which is used to supplement or correct the primary judgment result.

[0090] "Model update processing" refers to the process of retraining, fine-tuning, or updating parameters of the trained model and prompt statement generation rules based on recently accumulated feedback information, achievement judgment results, and historical information.

[0091] "Prompt statement generation rules" refers to a set of predefined logic, templates, or algorithms used to automatically construct prompt statements based on input conditions, target parameters, and context information.

[0092] "Learning data" refers to a collection of data with inputs and expected outputs or evaluation labels used to train or update machine learning models and generative artificial intelligence models.

[0093] In a preferred embodiment of the present invention, the server is configured to operate as a central processing node, and the terminal is configured to operate as a user interaction node, allowing the user to interact with the server via the terminal. The system of the present invention can be deployed in a data center, cloud computing platform, or enterprise intranet environment. The server can employ general-purpose computer hardware, such as a multi-core central processing unit, graphics processing unit, main memory, solid-state storage device, and network interface device. The server can run a general-purpose operating system, such as a UNIX-like operating system, and a backend application framework, such as a web server program built based on an interpreted or compiled language. The terminal can be a desktop computing device, a portable computing device, or a mobile communication device, running a browser program or local applications for displaying a user interface and sending requests.

[0094] At the hardware level, the server includes: a central processing unit and a graphics processing unit for performing numerical calculations and neural network inference; main memory for caching model parameters and intermediate features; non-volatile storage for persistently storing historical information and model files; and a network interface card for exchanging data with terminals. At the software level, the server includes: logical components such as a network communication module, a data preprocessing module, a feature extraction module, a machine learning model inference module, a generative artificial intelligence model invocation module, a sentiment analysis module, a visualization generation module, and a model update module. These modules can work collaboratively through inter-process communication or function calls.

[0095] When implementing machine learning models, servers can employ neural network architectures based on multilayer perceptrons, convolutional neural networks, or transformer structures, and leverage deep learning frameworks (such as frameworks supporting tensor operations and automatic differentiation) for training and inference. During training, the server uses historical information as training data and updates the weight parameters of each layer through forward and backward propagation algorithms. The server can use mean squared error, cross-entropy, or other loss functions as error metrics and employ stochastic gradient descent, momentum methods, or adaptive learning rate optimization algorithms for weight updates. During the training phase, the server can employ data augmentation methods, such as synonym replacement, sentence rearrangement, or noise injection, to improve the model's robustness and generalization ability. During the inference phase, the server loads the trained parameters to efficiently calculate target parameters and determine outputs for new inputs.

[0096] When representing data structures, the server can construct a record entity for each target request, which includes fields such as: request identifier, user identifier, original topic text, original hierarchical labels, standardized topic vector, hierarchical encoding, generated target parameters, generated structured target information, generated target text, user feedback content, achieved content, judgment result, and sentiment state. Internally, the server can convert topic information into word vectors or sentence vectors, for example, by using word embedding-based vectorization methods to map text to a fixed-dimensional real-valued vector space. When encoding hierarchical information, the server can use one-hot encoding or embedded vector representation to map different levels to a low-dimensional continuous vector space for neural network processing.

[0097] When generating structured target information, the server first retrieves similar records from historical data based on topic and hierarchical information, extracting features such as historical target values, achievement rates, and achievement times. Using a trained model, the server feeds the currently input topic and hierarchical vectors, along with historical statistical features, as input feature vectors into a neural network. The output layer of the neural network can be designed as regression output nodes to predict parameters such as the improvement rate and suggestion time period, or it can include classification output nodes to predict the target difficulty level or target type. After obtaining the neural network output, the server applies preset threshold and constraint conditions, such as limiting the improvement rate to a certain range and mapping the time period to a predetermined discrete set, thereby obtaining the target parameters within the business boundary.

[0098] When generating structured target information, the server can construct a unified data structure, including fields such as: target identifier, time period (e.g., "next quarter"), achievement indicator type (e.g., "sales revenue"), target value (e.g., percentage increase), benchmark value (e.g., current sales revenue benchmark), and target difficulty level. When converting these parameters into natural language, the server employs a generative artificial intelligence model, controlling the output content through prompts. When constructing prompts, the server combines topic information, hierarchical information, and target parameters in a clear textual format, for example: "You are a corporate performance management expert. Please generate a quarterly performance target for 'middle-level sales staff' based on the following information:" 1) Target direction: Increase sales revenue for the sales department; 2) Suggested increase: 10%; 3) Time frame: Next quarter; Please describe a specific target sentence in concise, clear, and quantifiable Chinese. The server provides the prompt as input text to the generative AI model. The generative AI model can employ a language model based on a transformer architecture, which includes multiple layers of self-attention encoding and decoding units. It utilizes word embeddings, positional encoding, and multi-head attention mechanisms to model the prompt. The server infers from the model and outputs the target text, such as: "Mid-level sales staff should increase their personal sales by 10% in the next quarter while maintaining current customer satisfaction." The server then performs post-processing on this target text, such as checking for sensitive terms, unreasonable numerical values, or structural errors, thereby filtering out inappropriate expressions.

[0099] When processing the achieved content, the server takes the actual execution results and related explanations submitted by the user as input. Users can input a text summary of their work or upload digitized data results via the terminal. The server performs sentence segmentation, word segmentation, and feature extraction on this text content to construct a representation vector for the achieved content. Simultaneously, the server also converts the previously generated target text or structured target information into feature vectors. When generating a prompt statement for a decision, the server combines the target information and the achieved content, for example: Please determine the achievement status of the objective based on the following information: Target: Mid-level sales staff will increase their individual sales by 10% in the next quarter.

[0100] Actual situation: This employee's sales increased by 9% this quarter compared to the previous quarter, and customer satisfaction rating remained above 4.5.

[0101] Please use one of the following words to make your judgment: 'achieved,' 'mostly achieved,' or 'not achieved,' and provide a brief reason. The server inputs the prompt into a generative artificial intelligence model to obtain a preliminary textual judgment on the degree of achievement. It can then map the output to discrete levels, such as through keyword matching or a classification sub-model, converting the natural language result into labels like "achieved," "basically achieved," or "not achieved." This approach allows the server to encode complex, unstructured descriptions of achievement into structured evaluation results through a language model, thereby enabling flexible judgment of the achieved content.

[0102] In terms of sentiment analysis, the server can be configured with independent speech analysis and image analysis models. When users provide feedback on the terminal, they can send facial expression images via voice input or a camera. The server extracts features from the speech signal, such as Mel-frequency cepstral coefficients, pitch, and energy, and inputs these features into the sentiment classification neural network. The server performs face detection, key point localization, and facial expression feature extraction on the facial expression images, converting facial muscle changes into feature vectors, which are then input into the facial expression sentiment recognition model. The server can fuse the speech sentiment results and facial expression sentiment results, for example, using a weighted average or multimodal fusion network, to obtain the user's sentiment state category. The server compares the sentiment state with a previous judgment result to see if there is a significant inconsistency. For example, if a judgment result is "achieved," but the sentiment state shows the user is clearly negative, the server can trigger rules to correct or add annotations to the judgment result, indicating information such as "the result was achieved, but the pressure is high." This technology does not simply imitate human psychological judgment, but dynamically adjusts the judgment result internally through multimodal feature fusion and rule logic.

[0103] When generating visualization output, the server converts target information and achievement determination results into the data structure required for graphics or charts, such as a set of data points for line charts, bar charts, or radar charts. The server constructs a list of sequential data based on a timeline and builds multi-dimensional data structures based on different achievement indicators. The display data generated by the server can include the target value, actual value, achievement rate, and corresponding sentiment status indicators for each period. The server sends this display data to the terminal, which then renders the corresponding images on the monitor. The terminal can perform rendering based on a web script library or a local graphics rendering library. Through this structured visualization data generation strategy, the server can allow the terminal to reproduce complex multi-dimensional information with fewer rendering instructions while maintaining data compression and transmission efficiency, thereby reducing communication load and improving overall response speed.

[0104] When processing user feedback and model updates, the server stores the feedback information along with the generated target text and achievement assessment results, and tags them with timestamps and associated users. The server can batch-extract these feedback records at predetermined time intervals to build new training datasets. For example, the server can add labels such as "user acceptance" and "satisfaction rating" to each target text to fine-tune the output style or constraints of the generative AI model. Simultaneously, the server can adjust the prompt generation rules based on common modification requests in the feedback, such as adding emphasis to phrases like "not too high" and "achievable." When retraining or fine-tuning the model, the server uses small learning rates and regularization strategies to avoid overfitting. Furthermore, the server can retain model parameters from different versions and roll back to a previous version when performance metrics are unsatisfactory. This continuous model update process allows the system to dynamically adapt to changes in organizational behavior patterns and the environment over time, improving the accuracy and stability of the output results at the computer system level.

[0105] The server employs a modular architecture to decouple functions such as data reception, preprocessing, target parameter calculation, prompt generation, generative AI model inference, sentiment analysis, visualization generation, and model updates. By maintaining feature extractors, word embedding matrices, model parameters, and rule configurations in memory, the server reduces repeated disk accesses, thereby improving processing speed. Under high concurrency, the server can use task queues and batch inference to merge multiple requests into batch tensors for matrix operations, improving graphics processing unit utilization and reducing model call overhead. Furthermore, the server compresses and structures the transmitted data, reducing the amount of communication data between the client and server, further minimizing network bandwidth consumption.

[0106] In implementing this invention, the terminal is responsible for providing the human-computer interaction interface and basic data verification. The terminal allows the user to input topic and hierarchy information through input controls, and performs simple format checks locally, such as mandatory field checks and length limit checks, to reduce invalid requests. After receiving the server's response, the terminal displays the target text and chart data to the user and provides feedback that the input area has been successfully completed. The terminal can reduce repeated requests to the server through caching technology; for example, when a user repeatedly views the same target within a short period, the terminal can directly read the display data from the cache. When transmitting voice and image data, the terminal can use compressed encoding formats to reduce network transmission load.

[0107] When using the system of this invention, users input organizational theme and hierarchical information via a terminal. For example, a user can input "increase sales revenue for the sales department" and select "middle-level employees" on the interface, then click the target generation button. After receiving the target text generated by the server, users can provide modification suggestions or feedback based on their own judgment, such as writing "Please appropriately reduce the target percentage" or "This target is reasonable." After completing a certain period of work, users can record the achievements on the terminal, such as "Sales revenue increased by 9% this quarter, and customer satisfaction remained at 4.5 points," and choose whether to include a voice explanation or turn on the camera so that the server can perform sentiment analysis. When viewing the achievement judgment results and charts generated by the system, users can intuitively understand the performance trends of themselves or their team and adjust their work strategies in a timely manner.

[0108] Through the aforementioned implementation, the server does not merely proceduralize manually set rules. Instead, it utilizes multi-layered neural networks, generative artificial intelligence models, sentiment analysis algorithms, and structured data management methods to implement a non-traditional goal setting and achievement determination process within the computer. Through adaptive prompt construction and model update mechanisms, the server establishes a closer statistical mapping relationship between the input and output spaces, thereby surpassing traditional rule-based systems in terms of technical effectiveness. This improves the accuracy and consistency of output results and achieves improvements in areas such as computing resource utilization, communication load, and data storage structure. These improvements represent optimizations of computer technology itself, rather than simply automating business processes.

[0109] use Figure 11 The processing procedure is explained.

[0110] Step 1: The user enters the target setting conditions on the terminal. On the terminal interface, users can enter the organization's topic information in the text input box, such as "sales department sales increase", and select the level information, such as "middle-level employees", through keyboard or touch operation.

[0111] The terminal takes the topic and hierarchy information entered by the user as input, performs basic format checks on it (such as whether it is empty or whether the length exceeds the limit), and then packages the data that passes the checks into the request data structure.

[0112] The terminal outputs the requested data structure and sends it to the server's designated interface address via HTTP / HTTPS protocol using the network communication module.

[0113] Step 2: The server receives and standardizes the input information. The server takes the request data from the terminal as input, receives message data containing subject and hierarchy information through the network interface, and the backend application parses the HTTP message to extract the subject and hierarchy fields.

[0114] The server performs string cleaning operations on the extracted topic information, such as removing leading and trailing spaces, standardizing encoding format, and filtering special characters, and compares and verifies the hierarchical information with the preset list of valid hierarchical levels.

[0115] The server uses a text preprocessing module to segment and tag the cleaned topic information, converting it into standardized lowercase, canonical terminology, and mapping the hierarchical information to internal hierarchical encoding, such as mapping it to "LEVEL_MIDDLE".

[0116] The server writes the standardized topic text, hierarchical encoding, and original input together into the storage device, generates a record containing the request identifier, and passes the standardization result as output to the subsequent target parameter calculation module.

[0117] Step 3: The server extracts historical information from the storage device and constructs features. The server takes standardized topic information and hierarchical coding as input and retrieves historical records similar to the topic and hierarchical level from the storage device, including historical target text, historical achievement rate, time period, and related indicator data.

[0118] The server cleans the retrieved historical information, deletes records with missing key fields, and marks or removes outliers (such as obviously incorrect values).

[0119] The server uses a feature extraction module to convert historical target text into vectors through word embedding or TF-IDF methods, and normalizes numerical achievement rates, time periods, etc., into numerical features within a unified range.

[0120] The server combines the current request's topic vector, hierarchical encoding vector, and historical statistical features (average achievement rate, typical improvement ratio, etc.) into a set of feature vectors, which are then used as output for the trained model to calculate target parameters.

[0121] Step 4: The server calls the trained model to calculate the target parameters. The server takes the feature vector generated in step 3 as input and feeds it into the trained neural network model deployed in the server's memory.

[0122] The server performs forward propagation computation of the neural network on the central processing unit or graphics processing unit, performing multi-layer linear transformations, non-linear activations, and possible attention operations on the input feature vector to obtain the predicted value of the output layer.

[0123] The server reads target parameters from the output layer, such as the suggested increase percentage (e.g., 10%), the suggested time period (e.g., "next quarter"), and the target difficulty level, and adjusts these parameters according to pre-set threshold conditions and constraints, such as truncating excessively high percentages within a reasonable range.

[0124] The server assembles the corrected target parameters into structured target information, including period information, type of achievement indicator, and target value, and provides this structured target information as output to the prompt statement generation module.

[0125] Step 5: The server generates a prompt statement based on the target parameters. The server takes structured target information, standardized topic information, and hierarchical information as input to construct prompts for the generative artificial intelligence model.

[0126] The server uses string template concatenation and conditional insertion to combine information such as target direction, suggested improvement amount, and time range into natural language prompt text, for example: "You are a corporate performance management expert. Please generate a quarterly performance target for 'middle-level sales staff' based on the following information:" 1) Target direction: Increase sales revenue for the sales department; 2) Suggested increase: 10%; 3) Time frame: Next quarter; Please describe a specific target sentence in concise, clear, and quantifiable Chinese. The server outputs the completed prompt statement to the generative artificial intelligence model calling module for subsequent natural language target text generation.

[0127] Step 6: The server invokes a generative artificial intelligence model to generate the target text. The server takes the prompt from step 5 as input, calls the generative artificial intelligence model through the internally deployed language model or an external interface, and sends the prompt to the model.

[0128] The server-controlled generative AI model performs multi-layer self-attention and decoding operations under the transformer architecture, predicting the output word by word based on the conditions in the prompt statement to obtain the target text in natural language form, such as: "Mid-level sales staff should increase their personal sales by 10% in the next quarter while maintaining existing customer satisfaction." The server performs post-processing on the target text output by the model, including character encoding correction, sentence integrity checks, and filtering of sensitive words or inappropriate expressions.

[0129] The server will use the post-processed target text as output and pass it to the display control module and storage module for display and subsequent feedback association.

[0130] Step 7: The server sends the target text to the terminal and records the generated result. The server takes the target text and structured target information as input and assembles them into a response data packet containing the target content, time period, achievement indicators, and request identifier.

[0131] The server sends the response data packet to the terminal through the network interface and writes the generated result record for the request into the storage device, associating the target text with the corresponding target parameters, prompt statements, and timestamps.

[0132] The server outputs the sending status of the response data packet for the terminal to receive and display.

[0133] Step 8: The terminal displays the target text and receives user feedback. The terminal takes the response data packet returned by the server as input and parses the target text, time period, and achievement indicator fields.

[0134] The terminal displays the target content on the screen in the form of text areas and visual elements, and provides feedback input controls (such as text boxes, rating bars, and buttons) so that users can read the target and input their evaluation or suggestions for improvement.

[0135] The terminal packages the evaluation information or correction request entered by the user in the feedback control together with the original request identifier into a feedback data structure, and sends it as output to the server over the network.

[0136] Step 9: The server stores user feedback and updates learning data. The server takes the feedback data structure sent by the terminal as input and parses out the feedback text, satisfaction rating, and associated request identifier.

[0137] The server locates the corresponding target generation record in the storage device, writes the feedback text, score, and timestamp into the fields associated with that record, and marks it as data that can be used for training.

[0138] The server adds the new feedback data to the learning dataset, updates the data index used for model fine-tuning and adjusting the rules for generating prompt statements, and outputs the updated learning dataset for the model update module to use in subsequent cycles.

[0139] Step 10: The user inputs the desired content and submits a judgment request. After the target execution period ends, users input the actual achievements through the terminal, such as entering "Sales revenue increased by 9% compared to the previous quarter, and customer satisfaction remained at 4.5 points" in the text area.

[0140] Users can choose to submit a completion determination request in the terminal, and optionally record a voice description via microphone or provide an emoji via camera.

[0141] The terminal packages the content text, accompanying voice data and image data, along with the target request identifier, into a decision request data structure, and sends it as output to the server.

[0142] Step 11: The server generates a prompt statement for a decision and calls the generative artificial intelligence model. The server takes the target content, the corresponding target text or structured target information as input, reads the original target from the storage device, and combines the two.

[0143] The server constructs a prompt statement for a single judgment based on a preset template, for example: Please determine the achievement status of the objective based on the following information: Target: Mid-level sales staff will increase their individual sales by 10% in the next quarter.

[0144] Actual situation: This employee's sales increased by 9% this quarter compared to the previous quarter, and customer satisfaction rating remained above 4.5.

[0145] Please use one of the following words to make your judgment: 'achieved,' 'mostly achieved,' or 'not achieved,' and provide a brief reason. The server outputs a prompt statement as the result of the decision, inputs it into the generative artificial intelligence model, and obtains a natural language description of the decision result. This description can then be mapped to discrete decision labels (such as "basically achieved").

[0146] Step 12: The server analyzes the user's voice and facial expressions and infers their emotional state. The server takes the user-uploaded voice data and image data as input and sends them to the voice feature extraction module and the image feature extraction module, respectively.

[0147] The server performs preprocessing on the speech data (noise reduction, framing, windowing) and extracts feature vectors such as Mel frequency cepstral coefficients, pitch, and energy; it performs face detection and key point localization on the image data and extracts expression feature vectors.

[0148] The server inputs voice features and facial expression features into the emotion classification model, performs forward propagation calculations, outputs a probability distribution representing the emotion category, and selects the emotion state label corresponding to the highest probability, such as "positive", "neutral" or "negative".

[0149] The server outputs the emotional state label and its confidence level to the achievement result fusion module, which combines it with the initial judgment result.

[0150] Step 13: The server integrates the initial judgment result with the emotional state generation achievement judgment result. The server takes the initial judgment label and sentiment status label as input, and performs judgment processing by combining them with preset fusion rules or a two-level decision model.

[0151] For example, if the server determines that the achievement is "achieved" and the emotional state is "severely negative", it will adjust the final achievement determination to "achieved (high stress)"; or if the server determines that the achievement is "basically achieved" and the emotional state is "positive", it will improve the result to "basically achieved (good trend)".

[0152] The server represents the merged achievement determination results in a structured form, including the main determination label, auxiliary descriptions, and sentiment annotations. This structured result is then used as output by the visualization generation module and the storage module.

[0153] Step 14: The server generates visual display data and sends it to the terminal. The server takes the structured achievement determination results and corresponding target information as input to construct data structures for graphical display, such as building achievement rate sequences by time dimension and multi-dimensional arrays by indicator dimension.

[0154] The server calculates the normalized coordinates and annotations of each data point based on the chart type (line chart, bar chart, radar chart, etc.) and packages them into a display data package.

[0155] The server will send the data packet and the judgment result in text form to the terminal, and the generated visual data structure will be used as output for the terminal to render.

[0156] Step 15: The terminal displays the achievement results and supports user viewing and feedback. The terminal takes the display data packet and judgment text returned by the server as input, calls the local graphics drawing module to draw the corresponding graphics on the display screen, and displays explanatory information such as "achievement judgment result" and "emotional state" in the text area.

[0157] Users can view goals, actual values, and achievement curves for different periods on the terminal interface, as well as refer to emotional status indicators, to intuitively grasp the performance trend.

[0158] When needed, the terminal allows users to provide further feedback on the judgment results and packages this feedback into a new feedback data structure as output to the server to further enrich the data required for model updates.

[0159] 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".

[0160] In modern production and business management scenarios, computer systems typically determine goals and their achievement levels using only fixed thresholds or simple rules, which presents the following technical problems: (1) In the target setting stage, computer systems often set targets based solely on human experience or static configuration, failing to make full use of historical business performance data and quality-related data stored in the database. This results in a lack of data-driven objectivity and refinement in target setting, and computing resources are not effectively used for predictive computing. (2) In traditional systems, generative artificial intelligence models are usually used only as simple text generation tools and are not tightly coupled with the structured data calculation process. They lack an integrated technical solution to transform “numerical prediction results” into “high-value explanatory text that can be directly used by human-computer interaction interfaces” and cannot form a unified data process from feature extraction, model operation to explanatory generation within the server. (3) In the achievement determination stage, the server generally only compares the target value with the actual value. It lacks the ability to comprehensively model the achievement indicators, the current business content and the historical evaluation results. It cannot automatically construct prompt statements for the evaluation task through generative artificial intelligence models, and thus cannot efficiently generate structured judgment results and explanatory texts inside the computer. (4) Existing emotion recognition and achievement determination are often independent of each other. The emotion analysis results are not formally incorporated into the decision logic of the server side. There is a lack of a supplementary determination mechanism that standardizes emotion indicators and integrates them with a single determination result in a programmatic manner. As a result, it is difficult to make multi-dimensional corrections to the achievement within a unified algorithm framework. (5) During the results display stage, the server usually only outputs simple report data. The front-end device is passively responsible for graphics rendering and lacks the ability to generate visualization data groups and display control signals on the server side. It cannot guarantee consistent visualization performance and reusable visualization logic between different terminals from the system architecture level.

[0161] Therefore, it is necessary to provide a new computer implementation scheme that, through the collaboration between data processing programs, statistical learning programs, and generative artificial intelligence models within the server, completes target numerical prediction, automatic construction of prompt statements, generation of text descriptions, determination of sentiment supplementation, and generation of visual control signals. This improves the automation and interpretability of the target setting and achievement evaluation process within a unified system architecture, and enhances the overall computing process and resource utilization efficiency of the server when handling the above tasks.

[0162] 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.

[0163] In this invention, the server includes a target value prediction unit for acquiring business unit setting theme information and business difficulty differentiation information, and combining them with historical business performance data and quality-related data to form a historical data set; a data preprocessing function for converting the historical data set into tabular data using data processing and statistical learning programs, and performing feature extraction, missing value handling, and indicator calculation; a model operation function for constructing a regression operation model based on the preprocessed data and inputting the features representing the business unit setting theme and business difficulty differentiation into the regression operation model to predict the target value; and a prompt message for automatically generating input for a generative artificial intelligence model based on the predicted target value and the historical data set. The system comprises a target description generation unit, which inputs the prompt statements into a generative artificial intelligence model to generate explanatory text including the reasons for the target setting and operational guidelines; a primary judgment unit, which acquires business achievement content data input through the terminal, compares the business achievement content data with the target value to calculate the achievement index, constructs an evaluation data set and corresponding prompt statements based on this, inputs the prompt statements into the generative artificial intelligence model to generate an initial evaluation text, and outputs an initial achievement judgment result; and a supplementary judgment unit, which uses an emotion analysis program to perform emotion recognition on voice and image information to obtain emotion indicators, and supplements and corrects the achievement based on the emotion indicators and the initial judgment result to generate a supplementary judgment result. This allows for an end-to-end computational process within the server, from historical data preprocessing, regression model prediction, automatic construction of prompt statements, generative artificial intelligence model text generation to emotion supplementation judgment and visual control signal output. This improves the objectivity and automation of target setting and achievement evaluation while optimizing computer resource management and data flow methods for the above processes, thereby improving the overall technical performance of the computer system in production and business management tasks.

[0164] "Information processing device" refers to an electronic device that includes at least one processor, memory and communication interface, and is used to execute program code and process input data. It may be a server device, a terminal device or other computing device.

[0165] A “processing unit” refers to a logical functional module implemented by a processor through the execution of program instructions. It is used to perform operations, judgments, and control on input data and can be implemented in software, hardware, or a combination of both.

[0166] "Business unit setting theme information" refers to identifying information that represents the goals, improvement priorities, or work policies pursued by a specific business department, production line, or task unit within a certain period. It can be text, coded, or structured data.

[0167] "Business difficulty differentiation information" refers to attribute information used to indicate the level or category of business tasks in terms of complexity, importance, or requirement level, which can be represented by level number, weight value, or category label.

[0168] "Historical business performance data" refers to quantitative or qualitative data related to business execution results recorded over a past period, including output quantity, efficiency indicators, cycle time, resource consumption, and other recorded information used to characterize business performance.

[0169] "Quality-related data" refers to recorded information used to characterize the quality of products or services, including defect rates, rework rates, defect types, quality inspection results, and related statistical indicators.

[0170] "Historical dataset" refers to a multidimensional dataset consisting of historical business performance data, quality-related data, and associated information such as time and object identifiers, which can be used for statistical analysis and model training.

[0171] "Data processing program" refers to a software program or script that runs on an information processing device and is used to read, transform, clean, aggregate, and perform other operations on raw data.

[0172] "Statistical learning program" refers to a software program used to perform statistical learning or machine learning processes such as feature construction, model training, model inference, and performance evaluation, including algorithm implementations for tasks such as regression, classification, and clustering.

[0173] "Tabular data" refers to a data structure organized in rows and columns, where each row represents a record and each column represents a field or feature, which facilitates statistical analysis and model calculations.

[0174] "Features" refer to numerical or symbolic representations extracted or constructed from raw data that can be used by models to learn the relationship between inputs and outputs, including continuous features, discrete features, or encoded categorical features.

[0175] "Missing value handling" refers to the process of completing or correcting null, unknown, or invalid values ​​in a dataset by deleting records, interpolating, filling in values, or using other methods.

[0176] "Indicator calculation" refers to the process of generating statistical indicators for analysis or modeling based on raw or preprocessed data through summation, averaging, ratio, variance, or other calculations.

[0177] "Regression model" refers to a mathematical or learning model used to predict continuous output variables based on input features, including linear regression models, nonlinear regression models, or other function mappings that aim to minimize prediction error.

[0178] "Target value" refers to a quantitative target used to guide business execution, which is predicted by a regression model based on the theme information and business difficulty differentiation information set by the business unit. Examples include production target value, defect rate target value, or efficiency target value.

[0179] The "target value prediction unit" refers to a functional module on an information processing device that calculates the input business-related features through data preprocessing and regression calculation models, thereby outputting the target value.

[0180] "Generative AI models" refer to AI models that can automatically generate text, speech, or other content based on input prompts or conditional information. They are typically trained using deep learning methods and can include language models or multimodal generative models.

[0181] "Prompt statements" refer to input text or structured instructions constructed to guide generative artificial intelligence models to perform specific generative tasks, including task descriptions, contextual information, constraints, and other content.

[0182] The “Target Description Generation Unit” is a functional module used to automatically construct prompt statements based on target values ​​and historical data sets, and call a generative artificial intelligence model to generate explanatory text containing the reasons for setting the target and operational instructions.

[0183] "Explanatory text" refers to text data output by a generative artificial intelligence model that uses natural language to describe the target numerical background, setting basis, implementation suggestions, and other content.

[0184] "Management storage unit" refers to storage resources used to store target values, explanatory text, judgment results and related metadata for subsequent querying, analysis and management. It can be a database, file storage or other storage media.

[0185] "Terminal device" refers to human-computer interaction equipment that allows users to input data and view results, including computer terminals, mobile terminals, industrial control terminals, and other devices that can communicate with servers.

[0186] "Business achievement data" refers to data information recorded by users or the system regarding the actual completion status after the target execution cycle ends, including actual output, actual quality indicators, execution process descriptions, and other data.

[0187] "Achievement indicators" refer to calculation results used to quantitatively evaluate the degree of difference between actual business results and target values, such as achievement rate, deviation value, or comprehensive score.

[0188] "Evaluation dataset" refers to a collection of data that is combined for the purpose of evaluating achievement, including target values, business achievement data, achievement indicators, and related contextual information.

[0189] "First-time evaluation text" refers to natural language text generated from evaluation datasets and output by a generative artificial intelligence model, used to describe the achievement status and preliminary evaluation opinions.

[0190] The “first-time judgment unit” refers to the functional module used to calculate achievement indicators, construct evaluation data groups and prompt statements, call the generative artificial intelligence model to generate first-time evaluation text, and determine the first-time achievement judgment result based on the text.

[0191] "A preliminary assessment result" refers to the initial assessment information on the achievement of a target, based on a single evaluation text and achievement indicators. This information may include a grade, label, or score.

[0192] "Emotion analysis program" refers to a software program or algorithm used to extract and classify features from input voice signals, images, or video data in order to identify the user's emotional state.

[0193] “Voice information” refers to digital sound data acquired through an audio acquisition device, which includes the user’s speech content and voice characteristics.

[0194] "Image information" refers to static or dynamic image data acquired through an image acquisition device, which includes the facial expressions or posture features of a user.

[0195] "Emotion recognition processing" refers to the process of analyzing voice information and / or image information to determine the category or intensity of a user's emotion.

[0196] "Emotion metrics" refer to quantitative or qualitative results that represent a user's emotional state, and can be emotion tags, emotion scores, or multidimensional emotion vectors.

[0197] The “supplementary judgment unit” refers to a functional module used to obtain emotion indicators using an emotion analysis program, and to combine the emotion indicators with the initial judgment result to correct or supplement the achievement judgment result.

[0198] "Supplemented achievement determination result" refers to the final or revised determination result of the goal achievement status obtained after considering emotional indicators and based on the processing of the supplementary determination unit.

[0199] "Past judgment history" refers to the collection of achievement judgment results, evaluation texts and related indicator data generated and stored in previous cycles, which are used for subsequent analysis and secondary evaluation.

[0200] "Secondary evaluation text" refers to natural language text output by a generative artificial intelligence model that provides a more comprehensive evaluation of the achievement based on the initial judgment result and the supplementary achievement judgment result.

[0201] "Final achievement determination" refers to the final conclusive determination of the achievement of the target formed in the secondary evaluation stage, which can be used for management decision-making and record keeping.

[0202] The "secondary judgment unit" refers to a functional module that generates prompt statements for secondary evaluation based on the primary judgment result, supplemented achievement information, and past judgment history, and calls a generative artificial intelligence model to generate secondary evaluation text and final achievement judgment.

[0203] "Data set for visualization" refers to a data set prepared for generating graphical displays, which includes target values, actual values, changes in achievement indicators, changes in quality-related indicators, and time information.

[0204] "Display control signals" refer to control parameters or commands generated by the server based on visualization data sets, used to instruct information display devices on what graphic elements, layouts, and styles to present.

[0205] "Chart format" refers to the presentation of data relationships through visual graphics such as bar charts, line charts, pie charts, radar charts, and dashboards.

[0206] "Information display device" refers to an output device used to present text, graphics or image information, including monitors, panels, projection devices or other visual display components.

[0207] The “Visualization Display Unit” refers to the functional module on the server side used to generate visualization data groups and display control signals, and drive the information display device to display the achievement evaluation results in the form of charts.

[0208] In one embodiment of the invention, the server is configured in the local area network of the production site. The server includes a multi-core general-purpose processor, main memory, non-volatile storage, and a network interface. The server stores an operating system, a relational database system, data processing programs, statistical learning programs, generative artificial intelligence model inference programs, and sentiment analysis programs in the non-volatile storage. The server loads these programs into multiple logical modules in the main memory to implement the functions of the units described in the claims.

[0209] In a typical configuration, the server uses a general-purpose operating system as its base platform. A scripting language runtime environment and data analysis libraries are installed on this platform for data reading, cleaning, and feature construction. Machine learning libraries are also installed on the same platform for building regression and decision models. Furthermore, a generative artificial intelligence model and an emotion recognition model based on a deep learning framework are deployed in a dedicated inference service. The server communicates with terminals via a network interface; these terminals can be industrial control terminals, mobile terminals, or desktop terminals.

[0210] When implementing the target numerical prediction unit, the server uses a relational database system to store historical business performance data and quality-related data. When reading this data, the server organizes each record into rows with fields such as timestamp, business unit identifier, production line identifier, business unit theme, business difficulty distinction, output, number of defective products, and downtime. The server manages these records in memory using a tabular data structure, with each column corresponding to a single field. The server performs deduplication, outlier detection and removal, and missing value imputation (e.g., using moving average or median) on the tabular data through data processing programs, and generates derived fields by calculating window averages, standard deviations, ratios, etc. The server thus obtains a set of feature columns for modeling, such as: encoded theme vectors, level values, average output over several periods, output volatility, and defect rate levels.

[0211] When constructing the regression model, the server uses a statistical learning procedure to combine the aforementioned features into an input matrix and uses outputs historically considered "reasonable targets" or expert-annotated target values ​​as output vectors. The server employs a linear regression algorithm as one implementation, where it internally minimizes the mean squared error loss function and obtains the weight parameters through analytical solutions or iterative optimization. Alternatively, the server can use a feedforward neural network regression model with a multi-layer fully connected structure. In this case, the server uses stochastic gradient descent or adaptive learning rate optimization algorithms to iteratively update the network weights, using mean squared error or smoothed L1 loss as the error function, and trains using a batch processing method. Through this training process, the server transforms the "reasonable target level" implicit in human experience into a computable function mapping.

[0212] When predicting target values ​​for the current period, the server encodes the theme and difficulty of the current business unit and combines this with dynamic data such as recent actual output and defect rate to construct a feature vector in the same format as during training. The server inputs this feature vector into a regression model to obtain predicted output or quality targets. The server can also apply constraints after prediction, such as ensuring the target value does not exceed a certain multiple of the historical highest safe production capacity or is not lower than a certain percentage of the historical average production capacity, to prevent the model from generating extreme values ​​when data distribution changes. In this way, the server automates target setting prediction at the numerical level, avoiding fluctuations caused by purely manual experience-based settings.

[0213] When implementing the target description generation unit, the server organizes the predicted target value and relevant historical data into text and structured information, and constructs prompts suitable for generative artificial intelligence models. In a specific example, the server generates the following prompt: "The department's theme is 'Production Efficiency Improvement,' and the production line's theme is 'A Line Efficiency Improvement,' with a level of 3. The average daily output over the past three months was 2000 units, with a defect rate of 2.0%. Based on the regression model's predictions, the suggested target daily output is 2200 units, with a defect rate control target of 1.8%. Please explain in concise and professional language why the target is set at 2200 units / day, and provide one or two risk warnings." The server sends the prompt statement to the language model inference engine via a generative artificial intelligence model inference program. In this embodiment, the server uses a neural network language model based on a multi-layered self-attention structure. The language model encodes contextual information in the prompt statement and internally calculates the dependencies between different fields using attention weights, thereby generating coherent explanatory text. The server receives the generated result, performs length control and sensitive content filtering, and stores the generated explanatory text along with the target value in a management storage unit, while simultaneously sending it to the terminal in text format.

[0214] When displaying targets and explanations, the terminal presents the target values ​​and explanatory text provided by the server on the human-machine interface. For example, the terminal displays "Today's target output: 2200 units, target defect rate: ≤1.8%" on the production line operation interface, and below it displays an explanation generated by a generative artificial intelligence model: "Increasing the daily output target from 2000 units to 2200 units can achieve an efficiency improvement of approximately 10% while maintaining the defect rate without increasing it..." This allows users to grasp the basis for target setting without understanding the internal parameters of complex statistical models.

[0215] After the production cycle is completed, users input the business completion data through the terminal. Users fill in information such as actual output, actual defect rate, and major failure times on the interface, and can also input free text descriptions, such as "Optimize changeover process to reduce changeover time." After performing basic format validation on the input, the terminal sends the data to the server through the communication interface.

[0216] When implementing a judgment unit, the server compares the received business achievement data with the target value, calculating achievement indicators such as achievement rate, target deviation percentage, and defect rate deviation. The server then combines the target value, actual value, achievement indicators, and user description text into an evaluation data set, and constructs a new prompt statement based on this data set. In a specific example, the server generates the following prompt statement: "The target for this production line was: a daily output of 2,200 units with a defect rate of no more than 1.8%. The actual result was: a daily output of 2,150 units with a defect rate of 1.6%. Based on these data, please conduct a preliminary evaluation of the achievement of the target and provide 2 to 3 specific improvement suggestions." The server inputs the prompt statement into a generative artificial intelligence model to obtain an evaluation text, such as "Although the target of 2200 items was not fully achieved this period, the performance in defect rate control was better than the target...". Internally, the server maps this text to a judgment result, such as "close to achieving the target," according to preset rules. In this process, the server can use methods based on keyword weights, sentiment analysis, or simple classifiers to convert natural language evaluations into structured labels.

[0217] When implementing the supplementary judgment unit, the server uses an emotion analysis program to process the voice and image information collected from the terminal. The server extracts features such as voice pitch, speech rate, and energy from the voice path, and facial expression key points and expression classification vectors from the image path. It then uses convolutional neural networks, recurrent neural networks, or multimodal fusion networks to classify these features into emotions, outputting emotion indicators representing emotional states such as tension, anxiety, and satisfaction. The server combines these emotion indicators with the initial judgment result and corrects the achievement judgment under preset rules. For example, when the data achievement boundary is ambiguous and the user is extremely tense, the server can internally mark the record as "requiring administrator review." By uniformly integrating emotion information on the server side, the system can introduce a new technical dimension without increasing the user's burden, improving the robustness of the judgment results.

[0218] When implementing the secondary judgment unit, the server constructs time-series data from the primary judgment result, the supplemented achievement judgment result, and past judgment history. The server can optionally use a time-series model or a memory-based neural network to model the trend of the judgment result over time, avoiding excessive influence of single-period anomalies on the final judgment. When generating secondary evaluation prompts, the server can provide richer context, such as: "The department's theme for this period was 'Production Efficiency Improvement,' at level 3. The target daily output was 2200 units, with a target defect rate of 1.8%. The actual daily output was 2150 units, with an actual defect rate of 1.6%. According to the rules, the achievement rate was 97.7%, and the initial assessment result was 'close to achievement.' The user added: 'Optimize the changeover process and reduce changeover time.' Please use a generative AI model to comprehensively evaluate the results of this period, explaining the main advantages and disadvantages, and providing three specific improvement suggestions." The server sends the prompt to the generative artificial intelligence model, which generates secondary evaluation text and a more stable final achievement judgment, and stores the results for further analysis.

[0219] When implementing the visualization display unit, the server pre-generates visualization data sets on the server side. The server aggregates target and actual values ​​according to the time dimension, forming a time-series array for drawing line charts; the server generates the data needed for bar charts based on production line or user dimensions; and the server calculates interval statistics based on achievement indicators to construct dashboards or radar charts. The server organizes this data into a structure consistent with the front-end chart library and generates display control signals, such as specifying chart type, axis range, and color scheme. Because the server centrally generates visualization data sets and display control signals, different terminals only need to execute simple rendering logic to present consistent results, thereby reducing the overhead of repetitive calculations and logic implementation on each terminal.

[0220] The technical effects achieved by the server through the above structure and processing methods include: automatically predicting target values ​​using regression models, thereby reducing the subjectivity of human settings; improving the accuracy of target prediction by characterizing historical data; achieving an integrated data flow from numerical calculation to natural language output internally by uniformly constructing prompt statements on the server side and driving generative artificial intelligence models to generate explanatory and evaluation texts, reducing the workload and inconsistencies of manual description writing; adding new technical dimensions and improving judgment robustness by introducing sentiment indicators and performing rule-based fusion on the server side without changing the front-end operation process; and reducing the repetitive calculation burden on each terminal by generating visualization data groups and display control signals on the server side, thereby improving overall processing speed and bandwidth utilization efficiency.

[0221] In various embodiments of this invention, the server can employ different model structures and algorithms. The server can use tree-based regression methods instead of linear regression to improve the fitting ability to nonlinear relationships; the server can use lightweight language models instead of large-scale language models to reduce inference latency and computational resource consumption; the server can use different types of emotion recognition models, such as voice-only or facial expression-only models, to adapt to different deployment environments. The server can also adjust the feature set and loss function design according to the business scenario, for example, assigning higher weights to extreme biases to enhance the model's sensitivity to anomalies. Through these alternatives, the system can adapt to the requirements of different hardware resources and application scenarios while maintaining its core structure, while continuously improving the accuracy and speed of target setting and achievement evaluation.

[0222] The terminal can be deployed in various ways depending on the implementation, such as as a touchscreen terminal installed in an industrial robot control cabinet, or as a tablet terminal used by workshop team leaders. Regardless of the terminal form, the core role of the server in this invention remains the same: processing and computing structured and unstructured data. Through the collaboration between regression models, generative artificial intelligence models, and sentiment analysis programs, it constitutes a computer system that goes beyond simple rule automation, but utilizes a novel computing architecture to improve overall technical performance. Users only input data and view results through the terminal, without directly participating in the model's internal calculations. Thus, the technical contribution of this invention is concentrated in the data flow organization and model coupling method on the server side.

[0223] use Figure 12 The processing procedure is explained.

[0224] Step 1: The server receives and stores basic configuration information.

[0225] Input: Department setting topic, production line topic, business difficulty classification (level), user identifier, etc. sent by the terminal.

[0226] The server parses the text and encoded data received from the terminal, verifies field integrity and data type, and writes fields such as "Department Setting Theme," "Production Line Theme," "Level," "Timestamp," and "User ID" into the setting information table in the database. During the write process, the server generates a task ID for this setting and stores it as one of the primary keys. Output: Setting record containing the task ID and structured setting data stored in the database.

[0227] Step 2: The server collects and preprocesses historical data.

[0228] Input: The task ID obtained in step 1, along with its associated department ID and production line ID.

[0229] Based on department and production line identifiers, the server reads records from historical business performance and quality data tables for a recent period (e.g., the past 6 months) including output, number of defective products, downtime, past target values, and achievement status. The server combines these records into tabular data, performing deduplication, outlier removal, and missing value imputation on each record. The server generates derived feature columns by calculating moving averages, standard deviations, maximum values, minimum values, and defect rate (number of defective products / total output). Output: The cleaned and expanded historical data table serves as a candidate feature dataset for model training and prediction.

[0230] Step 3: The server constructs feature quantities and establishes a regression calculation model.

[0231] Input: The historical data table output from step 2 and the coding scheme for setting topics and levels.

[0232] The server encodes the "department setting theme" and "production line theme" (e.g., one-hot encoding or integer encoding), and quantifies the "level." These feature columns are combined with statistical indicators such as historical output and defect rate to form a feature matrix. The server extracts "reasonable target values" from historical records, obtained through manual verification or rule derivation, as label vectors. The server calls a statistical learning program to train a linear regression model or a multilayer feedforward neural network model based on the feature matrix and label vectors. It iteratively updates the model parameters by minimizing the mean squared error or smoothing the L1 loss function until the loss converges or a predetermined number of iterations is reached. Output: The trained regression model and the encoding configuration used for feature processing, stored on the server's storage medium.

[0233] Step 4: The server predicts the target value for the current period.

[0234] Input: The current task's set theme and level stored in step 1, and the dynamic data such as the actual output and defect rate in the recent period (e.g., the last 7 days) in step 2.

[0235] The server encodes the current topic and level in the same way as in the training phase, and extracts recent statistical indicators as dynamic features. These features are combined into an input feature vector. The server calls the regression model obtained in step 3 to perform forward computation on the feature vector, obtaining predicted values ​​such as production targets and defect rate targets. The server then adjusts the predicted values ​​according to preset boundary rules (such as not exceeding a certain multiple of the historical highest safe production capacity). Output: Target value records associated with the task ID, including target production, target defect rate, etc.

[0236] Step 5: The server generates prompts to explain the target and calls a generative artificial intelligence model to generate explanatory text.

[0237] Input: The target value obtained in step 4 and the statistical information output in step 2 (such as historical average yield and historical defect rate).

[0238] The server organizes the target values ​​and key statistical data into natural language points, concatenating them into a prompt statement, such as: "The department's theme is 'Production Efficiency Improvement,' and the production line's theme is 'A Line Efficiency Improvement,' with a level of 3. The average daily output over the past three months was 2000 units, with a defect rate of 2.0%. Based on the regression model's prediction results, the suggested target daily output is 2200 units, and the defect rate control target is 1.8%. Please explain in concise and professional language why the target is set at 2200 units / day, and provide 1 to 2 risk warnings." The server inputs this prompt statement into the generative artificial intelligence model's inference interface, where the model's internal neural network encodes and decodes the word vectors, outputting a coherent explanatory text. Output: Target explanatory text corresponding to the target values, stored in an explanatory text table.

[0239] Step 6: The terminal acquires and displays the target value and explanatory text.

[0240] Input: Task ID, target value, and description text provided by the server.

[0241] The terminal requests target information for the currently active task from the server and receives response data including target output, target defect rate, and explanatory text. The terminal displays the target values ​​prominently on the interface, along with explanatory text below them. The terminal can map fields such as output targets and quality targets to designated controls for quick understanding by field operators. Output: A visually displayed target information interface for user review and confirmation.

[0242] Step 7: Users enter the business completion data after the cycle ends.

[0243] Input: The target information displayed on the terminal interface and the actual production results known to the user.

[0244] Users enter data such as actual daily output, number of qualified products, number of defective products, and downtime into the terminal interface, and can also input free text descriptions (e.g., "Optimize changeover process to reduce changeover time"). The terminal validates the user input for format and range (e.g., output is non-negative, defect rate does not exceed a reasonable upper limit), packages the validated data into an achievement content data package, and sends it to the server. Output: Structured achievement content data transmitted to the server over the network.

[0245] Step 8: The server calculates achievement metrics and organizes evaluation datasets.

[0246] Input: The target value record from step 4 and the achieved content data from step 7.

[0247] The server performs numerical calculations on the target output, target defect rate, actual output, and actual defect rate to obtain achievement indicators such as achievement rate, output deviation, and defect rate deviation. Based on time and task ID, the server combines these indicators with the original achievement data and target values ​​to form a unified evaluation data record, constituting the evaluation data cluster. The server adds a timestamp and user identifier to each evaluation record for easy subsequent retrieval and multi-period comparison. Output: A structured evaluation data cluster stored in the evaluation data table.

[0248] Step 9: The server generates prompts based on the evaluation data set and calls a generative artificial intelligence model to generate an evaluation text and a judgment result.

[0249] Input: Evaluation dataset generated in step 8.

[0250] The server extracts target values, actual values, achievement rates, and defect rate deviations from the evaluation data, and reads the user's free text description to construct a prompt statement, such as: "The target for this production line is: a daily output of 2200 units, with a defect rate not exceeding 1.8%. The actual result is: a daily output of 2150 units, with a defect rate of 1.6%. Based on this data, please provide a preliminary evaluation of the target achievement and offer 2 to 3 specific improvement suggestions." The server sends the prompt statement to the generative artificial intelligence model, which generates an evaluation text through inference in its internal encoding and decoding layers. Subsequently, the server maps the evaluation text to a judgment result (such as "achieved," "close to achieving," or "not achieved") according to preset rules (e.g., based on achievement rate thresholds and key terms in the evaluation text). Output: An evaluation text corresponding to the task ID and a judgment result.

[0251] Step 10: The server performs emotion recognition and generates emotion metrics.

[0252] Input: User voice and image information uploaded by the terminal.

[0253] The server inputs speech information into the speech feature extraction module to calculate parameters such as pitch, energy, and speech rate. It also inputs image information into the image processing module to extract facial key points and expression features. The server inputs these two modalities into the emotion recognition model. The model calculates the probability distribution of emotion categories through convolutional layers, recurrent layers, or self-attention layers. The server then selects or calculates emotion labels and emotion intensity scores as emotion indicators. Output: Emotion indicator records associated with the user and task ID.

[0254] Step 11: The server performs supplementary judgments based on sentiment indicators and the initial judgment result.

[0255] Input: The result of the judgment in step 9 and the sentiment index in step 10.

[0256] The server combines sentiment indicators with the initial judgment result based on preset rules or a lightweight judgment model. For example, when the achievement rate is close to a threshold and the sentiment indicator shows extremely negative sentiment, the server marks the judgment result as "requires review"; when the achievement rate is high and the sentiment is stable, the server directly confirms the initial judgment result as the supplementary judgment result. The server stores the supplementary judgment result together with the initial judgment result for subsequent secondary evaluation or trend analysis. Output: Supplementary achievement rate judgment result.

[0257] Step 12: The server generates a secondary evaluation prompt and calls a generative artificial intelligence model to generate the secondary evaluation text and the final achievement determination.

[0258] Input: initial judgment result, subsequent judgment result, and historical judgment records.

[0259] The server retrieves judgment results and achievement indicators from the historical judgment table, relating them to the current user or production line across multiple past periods. This historical trend information is combined with the initial judgment result and supplementary judgment result of the current period to form an extended evaluation context. Based on this context, the server generates a prompt statement for secondary evaluation, such as: "The department's theme for this period is 'Production Efficiency Improvement,' level 3. The target daily output is 2200 units, and the target defect rate is 1.8%. The actual daily output is 2150 units, and the actual defect rate is 1.6%. According to the rules, the achievement rate is 97.7%, and the initial judgment result is 'Close to Achievement.' User supplementary explanation: 'Optimize the changeover process to reduce changeover time.' Please use a generative AI model to comprehensively evaluate the results of this period, explaining the main advantages and disadvantages, and providing three specific improvement suggestions." The server calls the generative AI model to generate secondary evaluation text and, based on the overall evaluation tendency and historical trends, determines a more stable final achievement judgment. Output: Secondary evaluation text and final achievement judgment result.

[0260] Step 13: The server generates data sets for visualization and display control signals, and the terminal displays charts.

[0261] Input: Target numerical record, evaluation data set, primary and secondary judgment results, and historical data set.

[0262] The server combines target and actual values ​​into a time-series array based on a timeline, calculates indicators such as achievement rate and defect rate changes for each period, and generates numerical data for line charts, bar charts, and dashboards. The server organizes these values ​​into a visualization dataset and generates display control signals (including chart type, coordinate range, grouping method, etc.) based on preset chart templates. The terminal receives the visualization dataset and display control signals, draws corresponding charts in its local graphics library, and displays the target vs. actual comparison curve, achievement rate trend, and final achievement determination label to the user. Output: A visualization interface for achievement in various chart formats presented on the terminal display device.

[0263] 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.

[0264] 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."

[0265] Traditional goal management and performance evaluation technologies suffer from the following shortcomings: First, in computer-based performance management systems, goal setting often relies on human experience and lacks a unified hierarchical logic. The server side struggles to automatically generate structured goal information based on different business units and levels, resulting in inconsistent goal data granularity and semantic ambiguity stored in data storage devices, making it difficult for subsequent data analysis programs to utilize effectively. Second, existing systems typically only perform simple keyword matching or rule-based scoring on user-input content. The server-side data processing does not fully utilize the natural language understanding and generation capabilities of generative artificial intelligence models. It cannot automatically construct high-quality prompts within a unified computational flow, nor can it machine-processably integrate the rule-based quantitative results with the evaluation results output by the generative artificial intelligence model. Therefore, it is difficult to achieve stable and scalable comprehensive judgment at the algorithmic level. Third, in existing technologies, the emotion analysis module and performance assessment module are mostly implemented separately. Even if emotion recognition functionality exists, it is not deeply coupled with the server-side achievement assessment logic. This prevents the use of the user's emotional state reflected in their voice or facial expressions as structured features in the comprehensive scoring calculation within the same data processing flow. Consequently, the computer system fails to perform unified modeling and optimization across multimodal data dimensions. Fourth, existing systems primarily display achievement results as static reports or simple lists. The server side lacks the ability to uniformly aggregate and time-series process evaluation results from multiple periods, levels, and groups. This makes it difficult to form a visual data structure that can be used for subsequent automatic analysis and prediction, limiting the possibility of further algorithm optimization based on this data.

[0266] Therefore, how to build a system on the server side that is centered on a generative artificial intelligence model, can automatically generate prompts and integrate them with the results of rule-based data analysis, dynamically correct the comprehensive judgment by combining sentiment analysis results, and store and visualize the above multi-source data in a structured manner, thereby improving the automation, scalability and algorithm robustness of computer-based target management and performance evaluation processes, has become a technical problem to be solved in this field.

[0267] 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.

[0268] In this invention, the server includes: a data generation unit for automatically generating prompt statements for setting multi-level goals based on the set theme of the business unit, and generating structured target information based on the prompt statements using a generative artificial intelligence model; a data acquisition and storage unit for receiving achievement content input by the user through an information input device, converting natural language text into structured data, and storing the structured data together with identification information and time information in a relational data storage device; a rule-based analysis unit for reading multiple achievement contents from the relational data storage device, performing keyword extraction and statistical calculation using a data analysis program, and quantifying the degree of achievement according to predefined evaluation rules to obtain a judgment result; and a data generation unit for generating structured target information based on the achievement contents and... The system includes a model fusion unit that automatically constructs prompts for input into a generative artificial intelligence model based on the initial judgment result, sends the prompts to the model to obtain evaluation information, and fuses the initial judgment result with the evaluation information to generate a comprehensive judgment result of achievement level; an emotion correction unit that calls an emotion analysis engine to analyze the user's voice and / or facial expressions to obtain emotion features, and corrects the comprehensive judgment result based on the emotion features; and a visualization output unit that aggregates comprehensive judgment results from different periods, levels, and / or groups to generate visualized data for time-series comparison, level comparison, and group comparison, and outputs the data graphically or in chart form through an information display device. This allows for an integrated processing flow within the same computer system, encompassing goal setting, achievement content collection, rule-based calculation, generative artificial intelligence model evaluation, emotion information correction, and multi-dimensional visualization output. This enhances the server-side processing capabilities for natural language achievement content and multimodal user data, reduces manual intervention, improves the accuracy and consistency of achievement level judgment, and enhances the overall computer system's processing efficiency and scalability in goal management and performance evaluation tasks.

[0269] A "system" refers to a computer implementation scheme consisting of multiple functional units that cooperate with each other through communication to achieve functions such as goal setting, content processing, evaluation, and display.

[0270] "Information processing device" refers to an electronic computing device used to execute program code, receive, parse, process, and control the flow scheduling of input data, including but not limited to servers, computer terminals, or virtual computing resources.

[0271] A “business unit” refers to a functional entity in an organizational structure that has relatively independent responsibilities and objectives, including but not limited to departments, teams, or project groups.

[0272] "Setting a theme" refers to abstract descriptive information associated with a specific business cycle or business unit, used to constrain and guide the target content, including but not limited to strategic direction, work priorities, or evaluation indicators.

[0273] "Hierarchy" refers to the different levels in an organizational or job structure based on the scope of responsibilities and the degree of authority, including but not limited to company level, department level, team level, and individual level.

[0274] "Multi-level objectives" refers to a set of interrelated objectives set for different levels, used to decompose and align objectives from top to bottom or bottom to top within an organizational structure.

[0275] "Target information" refers to a set of data used to represent the items to be achieved and their evaluation criteria, including but not limited to target name, target content, quantitative indicators, time requirements, and weights.

[0276] "Generative artificial intelligence models" refer to algorithmic models trained based on machine learning and deep learning techniques, which are used to automatically generate text or other data based on input prompts and are capable of understanding and generating natural language.

[0277] "Prompt statements" refer to instructional or guiding text input into a generative artificial intelligence model, used to limit the model's output content, style, and format.

[0278] "Information input device" refers to a hardware or software interface used to receive user input data, including but not limited to graphical user interfaces, form pages, or application interfaces on terminal devices.

[0279] "Individual Achievements" refers to the actual work results and outcomes described by users for a specific period or goal, usually given in natural language text form.

[0280] “Text information” refers to content data represented in the form of a sequence of characters, which may include natural language sentences, numbers, symbols and their combinations.

[0281] "Structured data" refers to data that is organized and stored according to a predetermined data structure, including at least field names and corresponding data values, to facilitate querying, statistics and calculation by programs.

[0282] "Relational data storage device" refers to a storage system used to manage data in the form of a relational model, including but not limited to relational database management systems and their storage media.

[0283] A "data record" refers to a set of related field values ​​stored in a relational data storage device in the form of rows or tuples, used to represent an achievement or an evaluation result.

[0284] "Identification information" refers to data items used to uniquely distinguish objects, including but not limited to user identifiers, record identifiers, or target identifiers.

[0285] "Time information" refers to time attribute data associated with the content or evaluation results achieved, including but not limited to creation time, update time, or evaluation period.

[0286] "Data analysis software" refers to program components that run on information processing devices and are used to perform statistical, computational, and analytical processing on structured data, including but not limited to data processing libraries or analysis tools.

[0287] "Keyword extraction" refers to the process of automatically identifying important words or phrases that are representative or relevant to the evaluation from text information.

[0288] "Statistical processing" refers to the process of summarizing, calculating, and analyzing multiple data records, including but not limited to operations such as counting, summing, averaging, weighting, and normalization.

[0289] "Evaluation benchmark" refers to a set of predefined rules and standards used to quantitatively evaluate the achieved content, including but not limited to scoring rules, weight settings, and threshold conditions.

[0290] "Achievement rate" refers to a quantitative indicator that reflects the completion of a goal, usually expressed in the form of a score or grade.

[0291] "First-time judgment result" refers to the achievement result obtained by first quantitatively calculating the achievement of the content based on rule-based data analysis and evaluation benchmarks.

[0292] "Evaluation information" refers to the analysis results and evaluation content about the achieved content generated by the generative artificial intelligence model based on prompts, including at least one of the following: rating suggestions and textual descriptions.

[0293] "Comprehensive judgment result" refers to the final achievement evaluation result obtained by fusing the judgment result of a single judgment with the evaluation information output by the generative artificial intelligence model.

[0294] An “emotional analysis engine” refers to an algorithm module or software component used to analyze a user’s voice signals and / or image signals to identify their emotional state.

[0295] "User sentiment" refers to the emotional characteristics identified from a user's voice or facial expressions through a sentiment analysis engine, including but not limited to positive, negative, neutral, and their intensity.

[0296] "Sentiment features" refer to structured data representations extracted from user sentiment and used for subsequent calculations, including category labels, probability distributions, or numerical vectors.

[0297] "Emotion correction" refers to the process of adjusting or weighting the existing comprehensive judgment results based on the emotional characteristics identified by the information processing device.

[0298] "Information display device" refers to a device or interface used to output visual information to users, including but not limited to a display screen, graphical user interface, or web page front-end.

[0299] "Graphics or charts" refers to two-dimensional or three-dimensional representations of data, including but not limited to line charts, bar charts, pie charts, radar charts, and combinations thereof.

[0300] "Feedback information" refers to a set of information generated and displayed to users based on comprehensive judgment results, reflecting the achievement of goals and directions for improvement.

[0301] "Time series comparison" refers to a processing method that compares and analyzes the achievement or evaluation results at different points in time or during different periods.

[0302] "Hierarchical comparison" refers to a method of comparative analysis of goals or achievement levels at different organizational levels.

[0303] "Group comparison" refers to a processing method that compares and analyzes the evaluation results of different user groups or organizational units.

[0304] "Visualized data" refers to data structures that have been preprocessed and are suitable for presentation by graphics or charts, including datasets prepared for plotting time series, hierarchical, or group comparisons.

[0305] In various embodiments of this invention, the server runs the core program as an information processing device, the terminal serves as an information input and display device, and the user interacts with the server through the terminal, thereby forming a goal setting and achievement evaluation system based on a generative artificial intelligence model and a sentiment analysis engine. The program of this invention is implemented on general-purpose hardware and software environments, but through specific data structures, model structures, and computational processes, improvements are made to the computer's internal data management, computational efficiency, and evaluation accuracy.

[0306] I. System Hardware and Software Composition The server is implemented on one or more electronic computing devices, which include at least one central processing unit, semiconductor memory, non-volatile storage media, and a network interface. The server runs server-side applications on an operating system. These applications are implemented using an interpreted programming language, such as Python. Specific software components used by the server may include: - Web application framework: The server uses a web application framework to implement the HTTP interface, such as a lightweight Python-based web framework; - Database Management System: The server uses a relational database management system, such as an open-source relational database software, to implement a relational data storage device; - Data analysis library: The server uses a data analysis library, such as an open-source analysis library for tabular data processing, to perform keyword extraction, statistical calculations, and rule-based scoring; - Generative AI Model Interface: The server communicates with external generative AI model services via HTTP or remote procedure calls. The generative AI model can adopt a pre-trained language model based on the Transformer architecture. - Emotional Analysis Engine: The server uses an emotional analysis engine composed of a speech emotion recognition model and a facial expression recognition model. This engine can be deployed locally or in the cloud. - Log recording and monitoring module: The server uses a logging component to write prompts, model outputs, and error messages into the log storage for easy analysis and optimization.

[0307] The terminal can be a general-purpose information terminal, such as a desktop computer, laptop computer, smart terminal device, or tablet device. A general-purpose web browser is installed on the terminal to load HTML documents, CSS styles, and JavaScript scripts, and to serve as an information input and display device. Users input data using the terminal's keyboard, mouse, touchscreen, camera, and microphone.

[0308] II. Overall Program Structure and Data Structure Design The server stores program modules in a storage medium. These modules include a target generation module, an achievement content acquisition module, a rule-based analysis module, a generative artificial intelligence fusion module, a sentiment correction module, and a visualization output module. When executing the program, the server performs unified scheduling of the data flow between these modules.

[0309] The server defines multiple tables in a relational database. The server can use the following data structures: 1. Target Information Table The server stores structured data of multi-level targets in the target information table. The server can use the following field structures: target identifier, business unit identifier, hierarchy identifier, target content text, quantitative indicator field, weight field, creation time field, etc. Through this structure, the server can retrieve, aggregate, and update multi-level targets using a relational model.

[0310] 2. Contents to be achieved The server stores the user-submitted natural language text and its structured representation in the Achievement Content table. Fields may include: Achievement Content Identifier, User Identifier, Target Identifier, Achievement Content Text, Structured Feature Fields (such as keyword vectors and numerical features), Rule-based Scoring Fields, Generative AI Model Scoring Fields, Overall Scoring Fields, Sentiment Feature Fields, Creation Time, and Update Time. The server uses this table to progressively accumulate features from raw text.

[0311] 3. Logs and metadata tables The server records metadata in the log table, including the prompts used when calling the generative AI model, the model version number, the returned raw text, the parsed numerical scores, and the sentiment analysis engine version. This data structure allows the server to track and retrain the model's behavior, thereby improving the overall system's stability and interpretability.

[0312] Through the specific data structure design described above, the server can efficiently perform operations such as JOIN, GROUP BY, and time window aggregation at the relational database level, thereby improving computational efficiency and optimizing data management.

[0313] III. Implementation Forms of Target Setting and Prompt Statement Generation In the target generation module, the server generates prompt statements for multi-level target setting based on the business unit's defined theme and organizational hierarchy information. The server encodes the defined theme, historical target samples, and organizational structure description into a constrained text pattern, and then combines them into the prompt statements. Example prompt statements include: "You are a corporate goal-setting assistant. Based on the following business theme, please generate three quantifiable goals for the company, department, team, and individual levels respectively. Each goal must include quantifiable metrics and time requirements. Business theme: 'Improve online product sales conversion rate.' Please output in segments according to the hierarchy." The server sends these prompts to the generative AI model. In one implementation, the generative AI model can employ a pre-trained language model based on a Transformer encoder-decoder architecture, which includes multiple layers of self-attention layers, feedforward network layers, and positional encoding. When invoking the model, the server specifies model parameters such as the number of attention heads, hidden layer dimensions, maximum output length, and temperature parameters to control the diversity and stability of the output.

[0314] After receiving the model output, the server parses the target text generated by the model into structured target information. Using a rule-based parsing algorithm, the server extracts fields such as target level labels, indicator values, percentages, and time spans from the model output and populates them into the target information table. This structuring process transforms the natural language output into a data representation that can be directly used by the database and analysis modules, thereby improving the automation and accuracy of subsequent processing.

[0315] IV. Implementation Forms of Content Acquisition and Structured Transformation The terminal displays a list of goals and an input area for achievements in the user interface. Users describe the actual work results in natural language on the terminal, for example: "Project A was completed this quarter, launching two weeks ahead of schedule, and boosting sales of related products by 20%." The terminal assembles the user-input text with associated target identifiers and time information into a JSON object and sends it to the server via an HTTP request. Upon receiving the object, the server parses it and writes the text information into the content table. Simultaneously, the server invokes the text preprocessing module to perform word segmentation, stop word filtering, and lemmatization on the text.

[0316] During the structured transformation process, the server uses a data analysis library to extract keywords from the text columns. The server can use TF-IDF weights, rule dictionary matching, or sub-word segmentation to identify keywords such as "early," "online," "sales," "increase," and "20%" from the text. The server encodes the keywords into vector features, such as a multi-dimensional sparse vector or dense embedding vector, and stores these features in the structured feature field.

[0317] The server uses this conversion from natural language to feature vectors to enable the content to enter the rule-based analysis module and the generative artificial intelligence fusion module, achieving a unified numerical representation. This representation is stored in computer memory in vector and matrix form, facilitating batch processing using linear algebra operations and improving computational speed.

[0318] V. Implementation Forms of Rule-Based Analysis and Single-Instance Judgment In the rule-based analysis module, the server performs batch calculations on structured features. Based on predefined evaluation benchmarks, the server assigns weights to different types of keywords. For example, the server can assign positive weights to keywords indicating overachievement and negative weights to keywords indicating delays or incomplete completion. The server uses vector dot products or linear combinations to multiply each feature by its corresponding weight and sum them to obtain a rule score for one judgment.

[0319] The server uses a data analysis library to achieve the above linear combination through vectorized operations. Compared to iterative processing of each record, vectorized processing reduces interpreter overhead and the number of function calls, thereby significantly improving the computational efficiency of batch scoring.

[0320] During the initial judgment phase, the server can also calculate trend characteristics based on time information and historical data, such as the average achievement rate and fluctuation range over past multiple periods, and incorporate these characteristics into the rule-based scoring calculation model. By introducing time-series features, the server further improves its scoring accuracy and stability.

[0321] VI. Implementation Forms of Generative Artificial Intelligence Model Fusion In the generative AI fusion module, the server takes both the initial judgment result and the achieved content text as input to construct prompt statements for the generative AI model. Example prompt statements include: "As a performance evaluation expert, please review the employee's achievement of the target based on the following work completed this period and the preliminary score, and provide a score from 1 to 100, along with an evaluation description of no more than 200 words. Achievement: 'Completed Project A this quarter, launched two weeks ahead of schedule, and increased related product sales by 20%.' Preliminary score: 92 points. Please output the following format: Score: X, Evaluation: ……." The server inputs the aforementioned prompts into a generative artificial intelligence model. Internally, the model encodes the prompts, mapping the content text, numerical scores, and instruction information contained within to a high-dimensional vector space. The model calculates dependencies between words using a multi-head self-attention mechanism and sequentially generates output tags during the decoding phase. The server controls the stability of text generation by setting the generation length, decoding strategy (such as greedy search or bundle search), temperature value, and top-k threshold. This parameterized control balances repeatability and diversity, thereby improving the consistency and robustness of the overall evaluation results.

[0322] After receiving the model output, the server uses a text parser to extract the score and evaluation text. For example, when the output is "Score: 95, Evaluation: This employee significantly exceeded the target in this period's work...", the server uses a regular expression to identify "95" as the generative AI model score and uses the subsequent text as evaluation information. The server writes the score to the generative AI model score field and the evaluation text to the evaluation information field.

[0323] When making a comprehensive judgment, the server calculates a weighted average of the rule-based score and the generative AI model score according to predefined weights, or uses a non-linear fusion function (such as a fusion function based on logistic regression or a simple neural network) to generate a comprehensive score. Through this dual-channel structure of "rules + generative AI model", the server combines interpretable linear rules with a deep model with strong expressive power, achieving higher evaluation accuracy and better robustness than a single rule or a single model.

[0324] VII. Implementation Forms of Emotional Analysis and Emotional Correction While the user inputs content on the terminal, the terminal can capture the user's voice and facial image data through the camera and microphone. The terminal then sends the compressed or downsampled audio stream and video frames to the server. The server performs multimodal feature extraction in the sentiment analysis engine: - In the speech emotion model, the server extracts features such as Mel frequency cepstral coefficients, pitch, and energy from the audio signal and inputs them into a convolutional neural network or a recurrent neural network to output the probability of the emotion category. In the facial expression recognition model, the server performs face detection and alignment on image frames, inputs the face region into a residual network or a lightweight convolutional network, and outputs the probability distribution of facial expressions.

[0325] The server merges the emotion probability vectors of speech and facial expressions into a unified emotion feature vector, which is then written into the emotion feature field. The server adjusts the overall score based on these emotion features. For example, when a user is identified as being in a clearly negative or stressed state, the server can adjust the confidence level of the score or adaptively correct certain feature weights, thereby reducing the interference of extreme emotions on the model's input text. This multimodal correction mechanism reduces misjudgments caused by single text signals and improves the system's evaluation stability in complex situations.

[0326] VIII. Implementation Forms of Visualization Output and Data Aggregation In the visualization output module, the server aggregates and calculates the comprehensive judgment results across multiple periods, levels, and groups. Using SQL aggregation functions and a data analysis library, the server groups the comprehensive scores in the content table by time and organizational hierarchy, then generates time-series tables and hierarchical comparison tables. The server encapsulates these aggregation results into specialized visualization data structures, such as objects containing timestamp arrays and corresponding score arrays, for direct use by the front-end chart library.

[0327] The terminal receives the visualization data in the browser and uses a front-end visualization library to create line charts, bar charts, or radar charts. Users can intuitively view the time trend of the comprehensive score, the differences between different levels, and the comparison between different groups on the terminal. This data structure and aggregation method allows the server to generate multi-dimensional visualization data in a single calculation, reducing the communication load and processing latency caused by multiple queries and repeated calculations.

[0328] IX. Explanation of Technical Effects and Causal Relationship Through the combined structure of the aforementioned structured data design, rule-based analysis, generative artificial intelligence model fusion, and sentiment correction, the server makes this invention not merely a simple automation of manual operations, but rather an improvement on computer technology itself in the following aspects: 1. In terms of data management, the server uses specially designed target information tables and achievement content tables to unify natural language text, structured features, rule scoring, generative artificial intelligence model scoring, and sentiment features into a relational model, which facilitates efficient retrieval and batch processing, thereby improving data organization and access efficiency.

[0329] 2. In terms of computational efficiency, the server utilizes vectorized operations and batch processing to perform rule-based scoring on a large amount of content, significantly reducing processing time compared to traditional rule-by-rule matching. The invocation of the generative AI model reduces the frequency of API calls and network load through batch processing and templated prompts.

[0330] 3. Regarding evaluation accuracy, the server employs a fusion structure of rule-based scoring and generative AI model scoring, further incorporating sentiment features for correction. This multi-source feature fusion mechanism creates a non-traditional weighted structure within the model, reducing the impact of errors from single information sources and thus improving the overall accuracy and robustness of the scoring results.

[0331] 4. Regarding model training and updates, the server records prompts, model outputs, and final scores in a log table. Through offline analysis, the fusion model can be retrained or the weight parameters adjusted, thereby iteratively optimizing the internal algorithm structure of the computer without changing the business process.

[0332] 5. In terms of communication and rendering, the server generates chart-oriented visualization data through aggregation calculations, reducing the number of requests the front end needs to make to obtain data of different dimensions, thereby reducing communication overhead and the rendering burden on the terminal.

[0333] 10. Other forms of implementation Servers can employ different types of generative AI models in various implementations. For example, servers can use single-language models, multi-language models, or instruction-tuned models; they can deploy medium-sized models locally or invoke large models via cloud services. Servers can also select different Transformer layer depths and hidden layer dimensions based on resource availability, achieving a balance between accuracy and resource consumption.

[0334] In the rule-based analysis module, the server can replace or supplement linear rules with gradient boosting tree models or shallow neural networks to achieve more complex nonlinear relationship modeling. In the sentiment analysis module, the server can add additional modalities such as physiological signals or keyboard input rhythm to further enrich sentiment features.

[0335] In one implementation, the terminal can be a mobile terminal that uses built-in sensors to collect location data or usage status, and writes this data as additional features into the content table for subsequent more detailed analysis and modeling.

[0336] Through the various implementation forms described above, the server, terminal, and user work collaboratively in the system constituted by this invention, enabling goal setting, content input, rule-based analysis, generative artificial intelligence model evaluation, sentiment correction, and visualization output to be completed within the same computing framework. This achieves comprehensive optimization of data structure, algorithm flow, and model structure within the computer, resulting in technical effects such as improved processing speed, enhanced evaluation accuracy, and reduced communication load.

[0337] use Figure 13 The processing procedure is explained.

[0338] Step 1: The server generates multi-level target prompts based on the set topics and organizational hierarchy information of the business units.

[0339] Input: Business unit identifier, setting theme text, hierarchy list (e.g., company level, department level, team level, individual level), historical target sample (optional).

[0340] The server combines the set topics and hierarchical lists, inserting metadata such as business unit names, target periods, and the number of targets to be output into predefined prompt templates to form prompt statements for generative AI models. For example, the server generates the following prompt statement: "You are a corporate goal-setting assistant. Please generate three quantifiable goals for the company, department, team, and individual levels, respectively, based on the following business topic. Each goal must include quantifiable indicators and time requirements. Business topic: 'Improve online product sales conversion rate'. Please output in segments according to hierarchy." The server encodes the prompt statement into a text request and sends it to the generative artificial intelligence model service via a network interface.

[0341] Output: Prompt text for generative artificial intelligence models.

[0342] Step 2: The server invokes a generative artificial intelligence model to generate multi-level target text and performs structured parsing.

[0343] Input: The prompt text generated in step 1, and the configuration parameters of the generative artificial intelligence model (such as model name, maximum output length, and temperature value).

[0344] The server takes the prompt as input, packages the model parameters into a request body, and sends it to the generative AI model via HTTP or a remote procedure call (RPC) interface. Internally, the generative AI model encodes the prompt and generates output text containing multi-level objectives through multi-layered self-attention computation and decoding. Upon receiving this output text, the server uses regular expressions and a rule parser to segment the output by level, further extracting fields such as the objective name, quantifiable metrics (e.g., percentage values, quantity values), and time requirements (e.g., "within this quarter," "within one month") from each objective. The server then combines these parsed fields into structured records, generating the objective information data structure.

[0345] Output: A collection of target information objects containing multiple levels and multiple target records, which is written into the target information table.

[0346] Step 3: The terminal provides users with a target browsing and content input interface, and collects user input.

[0347] Input: The collection of target information objects returned by the server, and user identity information.

[0348] The terminal renders a page containing a list of goals and an input form via a web browser, displaying the goal content and related metrics for each level. Users select one or more goals in the terminal interface and enter the actual achievements in natural language in the text input boxes, such as: "Completed Project A this quarter, launched two weeks ahead of schedule, and increased related product sales by 20%." After the user clicks the "Submit" button, the terminal assembles the user-selected target identifier, the entered content text, the client time, and the user identifier into a data object. The terminal then serializes this object to generate a JSON or equivalent structured request body.

[0349] Output: Structured input data containing user identifier, target identifier, achieved content text, and time information, ready to be sent to the server.

[0350] Step 4: The terminal sends the completed content data to the server, which receives and stores the original record.

[0351] Input: The structured input data generated in step 3.

[0352] The terminal sends structured data to the server's specified interface URL via an HTTP POST request. Upon receiving the request, the server parses the request body, mapping the user identifier, target identifier, achievement content text, and time information to database fields. The server performs data validation (e.g., checking if the text is empty or if the identifier is valid). If validation passes, the server constructs an insert statement and adds an achievement content record to the relational data storage device.

[0353] After a successful write, the server retrieves the primary key identifier of the new record from the database to confirm that the completed content has been persistently saved.

[0354] Output: A new record of the achieved content in the database and its unique identifier, and a confirmation response returned to the terminal.

[0355] Step 5: The server preprocesses the content text and generates text features.

[0356] Input: The text of the achieved content, the corresponding target identifier, and the time information stored in the Achievement Content Table.

[0357] The server reads the completed content records that have not yet undergone feature processing, and performs a series of preprocessing operations on each text, including character normalization, whitespace cleanup, and sentence segmentation. The server then calls the word segmentation component to split the text into words or sub-words, removes common functional stop words, and uses predefined dictionaries or rule matching to identify performance-related keywords, such as "ahead of schedule," "delayed," "completed," "not completed," "growth," "decrease," and "percentage value."

[0358] The server converts the identified keywords into numerical features, for example, by generating feature vectors using a bag-of-words model or TF-IDF method, or by looking up pre-trained word vectors in a table. Multiple vectors are then combined into a sentence-level feature vector through averaging or concatenation. The server writes these features into the feature fields of the resulting content table for use by subsequent rule-based analysis and generative artificial intelligence fusion modules.

[0359] Output: A structured record of the achieved content with text feature vectors and keyword tags.

[0360] Step 6: The server performs rule-based scoring based on text features and evaluation benchmarks, generating a single judgment result.

[0361] Input: Text feature vectors generated in step 5, keyword tags, and predefined evaluation benchmarks (including keyword weights, thresholds, penalty terms, etc.).

[0362] The server assigns a corresponding weight to each keyword. For example, it assigns a positive weight to "complete ahead of schedule" and a negative weight to "delayed." For patterns with "growth of X%", it assigns a non-linear weight based on the value of X. The server multiplies the text feature vector with the weight vector using vector dot product or linear combination and then sums the results to obtain an initial score. The server further refines the score based on time information, target difficulty, or historical scores. For example, it increases rewards for maintaining a high achievement rate for multiple consecutive periods and adds penalties for significant fluctuations.

[0363] The server normalizes the obtained score to a predetermined range (e.g., 0-100), uses it as a judgment result, and writes it into the score field of the achievement content table.

[0364] Output: A judgment score for each achieved content, and an update record containing that score.

[0365] Step 7: The server constructs prompts for the generative artificial intelligence model and requests the model to provide auxiliary evaluation.

[0366] Input: Content to be achieved, a pre-calculated score, and target information (such as target content and indicator requirements).

[0367] The server embeds the above input into a structured prompt template, generating prompt statements for generative AI models. For example, the server generates the following prompt statement: "As a performance evaluation expert, please review the employee's achievement of the target based on the following work completed this period and the preliminary score, and provide a score from 1 to 100, along with an evaluation description of no more than 200 words. Achievement: 'Completed Project A this quarter, launched two weeks ahead of schedule, and increased related product sales by 20%.' Preliminary score: 92 points. Please output the following format: Score: X, Evaluation: ……." The server combines this prompt statement with the model parameters, packages it into a request, and sends it to the generative artificial intelligence model service via the network interface.

[0368] Output: A request for a prompt statement to be sent to the generative artificial intelligence model.

[0369] Step 8: The server receives the output of the generative artificial intelligence model and extracts evaluation information and model scores.

[0370] Input: The text response returned by the generative artificial intelligence model, which includes a score field and evaluative text.

[0371] The server analyzes the model output and performs parsing operations on the text. For example, the server uses regular expressions to find the number after "score:" and parses that number as the model score; the server treats the sentence after "evaluation:" as the text evaluation content. The server separates the model score and the evaluation text, removes extra whitespace and punctuation noise, and performs length truncation or sensitive information filtering on the text evaluation.

[0372] The server then writes the model score to the generative artificial intelligence model score field of the achievement content table and writes the evaluation text to the evaluation information field for subsequent comprehensive scoring and display.

[0373] Output: Generative AI model score and text evaluation information for each achieved content.

[0374] Step 9: The server combines the initial judgment score with the score from the generative artificial intelligence model to generate a comprehensive judgment result.

[0375] Input: initial judgment score, generative AI model score, optional historical scores, and target difficulty coefficient.

[0376] The server calculates a comprehensive score based on a preset fusion strategy. For example, the server calculates the score using a weighted average: Comprehensive Score = First Judgment Score × α + Generative AI Model Score × (1 / α) α), where α is a configurable weight. The server can dynamically adjust α based on the target category or historical consistency to reduce the impact of outliers on one side.

[0377] The server can also incorporate a simple two-layer neural network as a fusion module, taking the initial judgment score, the generative AI model score, and the historical stability index as input features, and outputting a corrected comprehensive score. When training this fusion network offline, the server uses mean squared error or cross-entropy as the loss function and updates the weight parameters through gradient descent, thereby making the comprehensive score closer to the manually labeled results.

[0378] The server writes the calculated comprehensive score into the comprehensive score field of the content achievement table, providing a basis for subsequent sentiment correction and visualization output.

[0379] Output: The overall score for each achieved item.

[0380] Step 10: The server performs sentiment analysis and corrects the overall judgment result based on sentiment features.

[0381] Input: User voice data, user facial image or video frame, and comprehensive judgment score.

[0382] During the user's input or submission of content, the terminal captures speech through a microphone and facial images through a camera, compressing this data before sending it to the server. The server then performs audio feature extraction (such as Mel-frequency cepstral coefficients, energy, and pitch) and image feature extraction (such as facial keypoint locations and expression-related textures) in its sentiment analysis engine, inputting these features into the sentiment recognition model. The model outputs a probability distribution of a set of sentiment labels, such as positive, negative, and neutral.

[0383] The server encodes this probability distribution into a sentiment feature vector and writes it into the content table. Based on pre-defined sentiment correction rules, if the server detects a user in an extremely negative state, it can reduce the penalty weight for negative expressions in the text or mark the record as requiring manual review; if it detects excessive self-aggrandizement, it can appropriately reduce the gain in the overall score. The server combines the sentiment features with the overall score using a linear or non-linear function to generate the final sentiment-corrected score.

[0384] Output: Includes the final achievement score of the emotional correction effect and the updated achievement record.

[0385] Step 11: The server aggregates scoring results from multiple periods, levels, and groups to generate visualized data.

[0386] Input: The final achievement score stored in the Achievement Content Table, the hierarchical information and time information in the Target Information Table.

[0387] The server selects relevant records from the database based on query conditions (such as user ID, business unit, and time interval), sorts the scores in chronological order, and constructs a time series. The server groups the scores by hierarchical fields, calculates the average score and distribution statistics for each hierarchical level, and performs aggregation calculations by group fields (such as department and project group) to obtain the group mean and variance. The server organizes these statistical results into a data structure suitable for plotting graphs, such as a collection of objects containing time arrays, corresponding score arrays, and hierarchical labels.

[0388] The server then returns the visualization data to the terminal in a compact JSON or other lightweight structure to minimize the amount of data transmitted over the network, thereby reducing the communication load.

[0389] Output: A collection of visual data objects used to create line charts, bar charts, or radar charts on the front end.

[0390] Step 12: The terminal displays visual results, allowing users to view and utilize feedback information for subsequent goal management.

[0391] Input: Visualized data generated in step 11, final achievement score, and text evaluation information.

[0392] The terminal uses a front-end charting library in the browser to render visualized data, plotting time-series scores as line charts and hierarchical or group comparison scores as bar charts or radar charts. Simultaneously, the terminal displays the final score for each record and evaluation text generated by a generative AI model, such as: "Score: 93. Evaluation: This employee significantly exceeded targets this period, completed the project ahead of schedule, and brought in a 20% increase in sales, demonstrating strong execution and business driving capabilities." Users view these graphical and textual information on the terminal screen, compare performance at different times and levels, and formulate work plans for the next cycle based on the strengths and improvement suggestions given in the evaluation. In this way, users make decisions using technical evaluation results calculated from the server, while the terminal helps users efficiently understand complex, multi-dimensional data through an interactive interface.

[0393] Output: A graphical display of achievement progress and text feedback information presented to the user, allowing the user to make subsequent operations and adjust goals accordingly.

[0394] 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".

[0395] In business scenarios centered on goal management and performance evaluation, existing computer-based evaluation systems typically suffer from the following technical problems: First, these systems often rely on fixed rules or simple threshold comparisons to calculate achievement levels. They struggle to automatically generate reasonable tiered goals based on different business themes, personnel levels, and historical data, resulting in a lack of flexibility and adaptability in goal setting. Computer resources passively execute pre-defined logic, failing to leverage the advantages of intelligent processing. Second, the achievement determination process is often disconnected from the natural language context of the business. The system only processes structured numerical values, struggling to effectively and promptly process unstructured data such as user-inputted text descriptions and on-site status information. This prevents the formation of a unified data representation and scalable reasoning mechanism within the computer. Third, existing systems generally ignore the impact of user emotional states on the understanding and acceptance of evaluation results. Even when emotion recognition is introduced, it is often limited to simple labeling or display, failing to tightly couple emotional data with achievement determination logic on the computer side and lacking dynamic correction and adaptive feedback mechanisms based on emotion. Fourth, for scenarios with high real-time requirements, such as production processes, traditional systems struggle to calculate real-time process achievement from multi-source sensor data and automatically generate targeted process improvement suggestions when the achievement falls below a threshold. This typically requires manual analysis, rendering the computer system merely a "access tool" in the data processing chain, rather than forming an end-to-end intelligent decision-making process. Fifth, existing visualization modules are mostly statically configured, unable to dynamically adjust the display style based on achievement results and emotional states. Computer-generated interfaces lack interactive design that aligns with users' psychological states, impacting the efficiency of indicator understanding and subsequent behavioral decisions.

[0396] Therefore, a new computer implementation scheme is needed to deeply integrate generative artificial intelligence models with sentiment analysis technology. By automatically constructing prompts on the server side and dynamically calling generative artificial intelligence models, a unified model is formed for goal setting, primary judgment, secondary judgment, and process improvement suggestion generation. Emotional states are integrated into the judgment and visualization control logic, thereby forming an adaptive, scalable, and real-time responsive intelligent evaluation and feedback processing flow within the computer to improve the overall system's intelligence, data processing efficiency, and human-computer interaction quality.

[0397] 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.

[0398] In this invention, the server includes a processing module for automatically generating prompt statements input to a generative artificial intelligence model based on the organization's business themes and the attribute information of the users, and generating graded business objectives based on the prompt statements and historical information; a processing module for acquiring business achievement content and quantity information from information terminals at the end of a cycle, calculating the achievement degree value internally using comparison and statistical operations, reconstructing prompt statements containing the comparison results and achievement content, and calling the generative artificial intelligence model to generate a first-order assessment result of the achievement degree and business improvement suggestions; and a processing module for receiving facial expression information and voice information acquired by image acquisition devices and voice acquisition devices, inferring the user's emotional state on the server side using the emotion analysis function, and correcting the achievement degree assessment based on the emotional state and the first-order assessment result, while generating a function to determine the emotion. The system includes a processing module that generates achievement assessment text and next-cycle target plans using a generative artificial intelligence model, along with a module for storing achievement assessments, business improvement suggestions, and next-cycle target plans in an information storage device, generating visualization data on the server side, embedding display information corresponding to the achievement assessment and emotional state within the visualization data, and outputting it to an information display device to drive dynamic presentation of graphics or charts. It also includes a processing module that continuously receives status and performance information from the running process, aggregates and statistically analyzes this information within a predetermined time window, calculates the real-time achievement of the process unit, and when the real-time achievement is lower than a benchmark value, generates a prompt statement containing status information and the real-time achievement, calls the generative artificial intelligence model to automatically generate process improvement suggestions, and outputs them to the display device or control device. This allows for the formation of a unified processing framework within the server, centered on generative artificial intelligence models and prompts. It deeply integrates target generation, primary judgment, secondary judgment, process improvement suggestion generation, sentiment analysis, and visual control, enabling the computer to adaptively adjust evaluation logic and display strategies for different business themes and emotional states. This achieves end-to-end automated processing from data collection and intelligent judgment to feedback presentation, thereby significantly improving the intelligence, real-time performance, and human-computer interaction experience of existing computer technologies in the fields of target management and performance evaluation.

[0399] A "system" refers to a collection of devices that are connected through communication, consisting of one or more processing devices, information storage devices, information display devices, and input / output devices, and are used to perform functions such as goal setting, achievement determination, sentiment analysis, visualization, and generation of improvement suggestions.

[0400] "Processing device" refers to an electronic device with computing and control functions, including one or more processors and the programs they execute, which are a combination of hardware and software used to analyze, calculate, judge and generate control or output data.

[0401] "Organizational business theme" refers to the business direction or task line determined by the organization during a predetermined evaluation cycle or project to guide goal setting and achievement assessment, such as abstract business goals like reducing defect rate, increasing production, or increasing sales.

[0402] "Attribute information of the user subject" refers to descriptive information related to the evaluated object, including but not limited to the department, job level, scope of responsibilities, historical performance level and business category, which are used to assist in generating characteristic data for graded business objectives and evaluation standards.

[0403] "Generative AI models" refer to AI models that can automatically generate text content or structured output results based on input prompts and contextual data, including deep learning-based language models or other models with content generation capabilities.

[0404] "Prompt statements" refer to text instructions or descriptions constructed by the processing device and input into the generative artificial intelligence model. They are used to describe the target task, constraints, and input data, enabling the generative artificial intelligence model to generate corresponding output statements or text fragments.

[0405] "Tiered business objectives" refers to a set of quantitative or qualitative business objectives that are differentiated and set for objects of different levels or categories based on the attribute information and business themes of different users.

[0406] "Business Achievement Content" refers to the actual execution results related to the established business objectives reported by the user or information terminal at the end of the target execution cycle, including numerical data, task completion status, and explanatory text.

[0407] "Quantitative information" refers to structured numerical data that can be used to calculate achievement, including indicators that can be mathematically calculated, such as output, sales, number of defective items, number of operations, and time consumption.

[0408] "Achievement numerical assessment" refers to the numerical result obtained by the processing device based on the comparison between business objectives and the content achieved in the business, through arithmetic operations, statistical operations, or model inference, to represent the degree of completion. This result may include percentages, scores, or standardized indicators.

[0409] "A first-time assessment result" refers to the achievement assessment result given by the processing device for a certain evaluation period or business unit based on the output of the generative artificial intelligence model, including at least one of quantitative scores and qualitative evaluation text.

[0410] "Business improvement suggestions" refer to actionable recommendations, optimization plans, or adjustment measures generated by a generative artificial intelligence model based on achievement assessment results and relevant contextual information, for the purpose of improving future business performance.

[0411] "Image acquisition device" refers to an imaging device used to acquire images of the subject's facial expressions or on-site scenes, including cameras, webcams, or other devices capable of outputting digital image data.

[0412] "Voice acquisition device" refers to an audio input device used to acquire the voice signal of the user, including microphones, microphones or other devices capable of acquiring and outputting digital audio data.

[0413] "Face information" refers to facial region or facial feature data obtained from an image acquisition device, used to reflect the emotional state of the user.

[0414] “Speech information” refers to the audio signal and its characteristic parameters obtained from the speech acquisition device, including pitch, volume, speech rate, and text extracted from the speech content, which are used to infer emotional state.

[0415] "Sentiment analysis" refers to programs or models that analyze facial and speech information to identify the emotional state of the user, including rule-based methods and models based on machine learning or deep learning.

[0416] "Emotional state" refers to the psychological or emotional category of the user at a certain moment, as identified by the emotion analysis function, such as joy, anger, frustration, neutrality, fatigue, etc.

[0417] "Achievement Assessment Text" refers to a natural language description generated or supplemented by a generative artificial intelligence model based on numerical achievement results and emotional state, used to describe the achievement status, cause analysis, and comprehensive evaluation.

[0418] "Next cycle target plan" refers to the target setting suggestions for the next evaluation cycle generated by the generative artificial intelligence model based on the current evaluation results. These suggestions include target values, priorities, and supporting explanations.

[0419] "Information storage device" refers to storage resources used to store information such as business objectives, achievements, evaluation results, emotional states, improvement suggestions, and visualized data, including disk storage, semiconductor storage, or network storage systems.

[0420] "Visualized data" refers to a structured collection of data generated for presentation in the form of graphics or charts on an information display device, which includes parameters for drawing coordinates, colors, shapes, text labels, and interactive behaviors.

[0421] "Display style information" refers to a set of parameters used to control the visual presentation style, including color, line type, fill style, emphasis, layout, etc., which are display control information corresponding to the degree of achievement and emotional state.

[0422] "Information display device" refers to an output device used to present visual data and text information to a user, including a display screen, touch screen, projection device or other terminal capable of displaying graphics and text.

[0423] "Running object process" refers to a business process or production process that is monitored and evaluated by the system, including a sequence of operations consisting of multiple steps or procedures.

[0424] "Status information" refers to data collected during the execution of a running object's process, used to describe the current running status of the process, including device operating status, queue length, interruption status, etc.

[0425] "Performance information" refers to data used to characterize the performance of a running object's process, including indicators such as output quantity, cycle time, resource utilization, and defect rate, which can be used for efficiency and quality analysis.

[0426] The “preset time window” refers to the time interval used for statistical analysis and aggregation of status and performance information. Its length and sliding method are preset or configurable in the system.

[0427] "Real-time achievement of process unit" refers to the immediate completion rate of a specific process unit within a predetermined time window, calculated by comparing actual performance information with target values.

[0428] "Benchmark value" refers to the threshold or reference value used to judge the level of real-time achievement. When the real-time achievement is lower than this value, improvement suggestions are generated or alarms are triggered.

[0429] "Process improvement suggestions" refer to actionable recommendations or adjustment schemes generated by generative artificial intelligence models to improve process performance when the real-time achievement rate of a running process or its process units is lower than the baseline value.

[0430] "Historical Business Improvement Implementation Information" refers to data recording the business improvement measures that were actually adopted and implemented in past evaluation cycles and their effects, which are used as a reference for subsequent judgment and target generation.

[0431] "Secondary assessment result of achievement" refers to the re-evaluation result of achievement output by a generative artificial intelligence model based on the initial assessment result, emotional state, and historical business improvement implementation information. It includes at least one of numerical and textual data.

[0432] "Final achievement assessment" refers to the comprehensive achievement evaluation result determined by the system after secondary judgment or correction processing, which is used for external display or archiving.

[0433] "Graphics or charts" refer to visual objects presented on information display devices to express achievement and other indicators, including line charts, bar charts, pie charts, radar charts, dashboard charts, etc.

[0434] In a preferred embodiment, the server is installed in a data center or factory server room. The server includes a multi-core central processing unit, a graphics processing unit, main memory, and non-volatile storage. The server runs a general-purpose operating system and has a scripting environment and a deep learning framework installed. The server stores in its storage device programs for executing the method of the present invention, parameters of the trained generative artificial intelligence model, parameters of the emotion recognition model, and visual configuration data.

[0435] In one implementation, the server includes a generative artificial intelligence model module, a sentiment analysis module, an achievement calculation module, a prompt statement generation module, a visualization data generation module, and a data storage module. The server uses a relational database or key-value database to store business topic information, user attribute information, goal setting records, achievement content records, achievement evaluation results, sentiment status records, and improvement suggestion records. The server interacts with terminals and production line control devices via communication interfaces.

[0436] In one embodiment, the terminal is a mobile information terminal or fixed terminal equipped with a display screen, a touch input unit, a camera pixel sensor, and a microphone sound sensor. The terminal runs a mobile operating system or a desktop operating system and runs browser applications or native applications. The terminal communicates with the server via a secure communication protocol. The user inputs the business topic, attribute information, and desired content through the terminal's graphical user interface, and provides facial expressions and voice signals to the server through the terminal's camera and microphone.

[0437] In one implementation, the server uses a generative AI model module to achieve prompt-driven target generation and evaluation. The generative AI model module employs a multi-layer transformer structure, including a multi-head self-attention layer, a feedforward neural network layer, layer normalization units, and positional encoding units. During the model training phase, the server uses large-scale text data and structured business data for pre-training and fine-tuning. During training, the server uses a cross-entropy loss function or a sequence-to-sequence loss function and updates the model parameters through backpropagation and adaptive learning rate optimization algorithms.

[0438] In one implementation, the server uses a prompt generation module to convert structured data into a natural language task description. When generating the target, the server constructs prompts based on the organization's business themes and the attributes of the users. The server embeds data such as business themes, levels, and historical performance characteristics as placeholders into template text, thus forming the prompts input into the generative AI model. For example, the server generates the following prompts: "Based on the department theme 'Reducing Production Defect Rate', generate quantitative targets for the next quarter for Level A operators. Please provide recommended values ​​for the upper limit of the defect rate and the number of daily inspections, and output the results in Simplified Chinese." After receiving the text output by the generative artificial intelligence model, the server uses a regular expression parsing module and a keyword extraction algorithm to extract numerical values ​​and constraints from the natural language into structured target data, such as "defect rate ≤ 1%, daily inspection ≥ 3 times". The server stores the extraction results in the database and uses them as target parameters for subsequent achievement calculations.

[0439] In one implementation, during achievement assessment, the server combines quantitative information (e.g., output, number of defective items, sales revenue, etc.) input by the user through the terminal with target parameters stored in the database to form a feature vector. The server standardizes these features in the achievement calculation module, subtracting the mean from each indicator and dividing by the standard deviation to eliminate dimensional differences. The server also calculates derived features, such as defect rate, output per unit time, and target deviation rate. Based on these features, the server uses a linear combination or a small feedforward neural network to obtain an initial numerical achievement score. For example, the server uses a single hidden layer neural network, with the hidden layer employing a rectified linear unit activation function and the output layer using a sigmoid function to constrain the result between 0 and 1.

[0440] In one implementation, after obtaining the numerical achievement level, the server generates prompts for the generative AI model, which include a judgment text and improvement suggestions. The server incorporates the target, actual results, and statistical features into the prompts. For example, the server generates the following prompts: "Target: Daily output of 500 units, defect rate ≤1%. Actual output: 450 units, 4 defective units, defect rate approximately 0.89%. Please evaluate the achievement in Simplified Chinese and provide three specific improvement suggestions that can be implemented within one week." The server inputs the aforementioned prompts along with numerical features into the generative AI model. The text output by the generative AI model includes an overall evaluation paragraph and an improvement suggestion paragraph. The server segments and annotates the output text, separating the evaluation section from the improvement suggestion section, and stores them in the database.

[0441] In one implementation, the server's sentiment analysis module consists of a facial expression recognition submodule and a speech sentiment analysis submodule. The server employs a convolutional neural network (CNN) structure for expression recognition, containing several convolutional layers, pooling layers, and fully connected layers. It takes the grayscale matrix of the facial image as input and outputs probability vectors for multiple sentiment categories. During deployment, the server stores this model as binary parameters in non-volatile storage. At runtime, the server uses a computer vision library to perform face detection and cropping on image data received from the terminal, then inputs the normalized image into the CNN to obtain the sentiment prediction result.

[0442] In speech sentiment analysis, the server extracts acoustic features such as short-time energy, fundamental frequency, formants, and Mel-frequency cepstral coefficients from the speech signal. The server uses recurrent neural networks or convolutional neural networks to model the feature sequence and outputs the sentiment category probability. The server then weights and fuses the facial expression sentiment probability with the speech sentiment probability, obtaining a comprehensive sentiment state label, such as "joy," "frustration," or "fatigue," through maximum probability rules or Bayesian posterior calculations.

[0443] In one implementation, after obtaining the initial judgment result and sentiment state, the server decides whether to perform a secondary judgment based on preset rules. The server can be configured to initiate a secondary judgment, for example, when the achievement level is below a certain threshold and the sentiment state is negative. In this case, the server generates a prompt statement containing the initial judgment result, sentiment state, and historical improvement implementation records. For example, the server generates the following prompt statement: "The user's achievement rate for this period is 60%, and their current emotional state is 'frustrated.' The improvement measures taken in the previous period included 'adjusting the frequency of material replenishment.' Please provide more encouraging evaluation text while maintaining the authenticity of the evaluation, and set two moderately difficult but quantifiable goals with room for improvement for the next period." The server inputs the aforementioned prompts into a generative AI model, extracting secondary evaluations and new target suggestions from the output text. Based on these secondary evaluations, the server can make minor adjustments to the initial numerical achievement level. For example, it can apply a non-linear mapping function to smooth out scores in the lower range, improving user acceptability while maintaining ranking stability.

[0444] In one implementation, the server's visualization data generation module generates graphical configuration data based on achievement assessment results and emotional states. The server constructs a data array for each dimension (e.g., output, defect rate, time-series achievement) and assigns visual attributes such as color, line style, and thickness. If the server detects an emotional state of "joy," it specifies a highlight color and a larger line width in the configuration data; if the emotional state is "frustration," it uses a softer or darker color scheme. The server sends this visualization data to the terminal in a structured format, and the terminal uses this data to draw line charts, bar charts, or gauge charts on the display screen.

[0445] In one implementation, the server receives status and performance information, such as production counts, cycle time, and defect counts, from the production line information processing unit. The server stores this data in a time-series table and applies a sliding window aggregation operation. The server uses, for example, a 5-minute window to calculate short-term average production and defect rate, and calculates real-time achievement based on target values. When the achievement falls below a preset baseline, the server immediately constructs a prompt statement containing the current window data, trend information, and target parameters. For example: "The current assembly process has achieved an average completion rate of 65% in the last 30 minutes, and the cycle time is 20% slower than planned. Please analyze the possible bottleneck factors and generate three process improvement suggestions for the assembly process, which should be implementable directly at the team level." The server inputs prompts into the generative artificial intelligence model to obtain process improvement suggestions for specific procedures. Based on these suggestions, the server can further generate structured rules, such as adjusting the work sequence, adding intermediate buffers, or modifying the inspection frequency. The server can then send these rules to the production control unit to automatically adjust certain equipment parameters, or push them to the shop floor terminal screens in the form of alarms and suggestions.

[0446] In one implementation, users view server-generated charts and text evaluations through a terminal interface. Users can mark adopted improvements on the terminal, and the server writes these marks as historical business improvement implementation information into the database. When generating prompts in the next round, the server inserts this historical improvement information into the prompts, enabling the generative AI model to propose new goals and measures based on existing improvements, thus forming a non-linear decision-making chain within the model that utilizes historical behavioral feedback.

[0447] In one implementation, the server achieves efficient utilization of computer resources through the aforementioned modular data flow. By unifying the complex goal setting, achievement calculation, sentiment correction, and process improvement suggestions into a unified abstraction of "prompt statement + feature data → generative AI model output," the server can reuse the same model structure to handle multiple tasks, thereby reducing storage redundancy and maintenance costs associated with modeling different tasks separately. When generating prompt statements, the server tailors input information according to the scenario, reducing the transmission and computation of irrelevant information each time the generative AI model is invoked, thus reducing communication and computational loads.

[0448] In another implementation, the server can deploy a smaller-scale generative AI model on an edge computing device for generating local process-level improvement suggestions, while deploying a large-scale model in the cloud for complex text analysis and strategy generation. The server can choose to invoke either the edge model or the cloud model based on network conditions and task complexity, thus ensuring both response speed and inference quality.

[0449] In one implementation, the server trains the sentiment analysis model using labeled facial expression images and speech samples through supervised learning. The server employs mean squared error or cross-entropy as the loss function and updates the weights of convolutional and fully connected layers using gradient descent-like algorithms. The server can introduce data augmentation methods, such as random cropping, micro-face rotation, brightness variations, and speech noise addition, to enhance the model's robustness to changes in real-world scenarios. By pre-training the model offline and performing only forward inference during the online phase, the server significantly reduces online computational costs and improves the overall system response speed.

[0450] In one implementation, the server encapsulates all modules as independent services, with modules exchanging data via message queues or remote procedure calls (RPCs). This modular architecture allows the server to horizontally scale modules based on load; for example, multiple server instances with graphics processing units can be configured separately for the generative AI model module, and a dedicated accelerator can be configured for the sentiment analysis module. Through this structure, the server as a whole can maintain low response latency and high throughput even under high concurrency requests.

[0451] Through the above implementation, the server internally uses prompts as an intermediary to uniformly map structured data, unstructured text data, sentiment data, and process data into the input space of the generative artificial intelligence model. This allows the model to perform multi-task reasoning within a unified framework, including goal setting, achievement evaluation, sentiment correction, and process improvement. Compared to simply automating manual calculations, this system, through specific network structures, feature engineering, and modular data flow design, achieves technical improvements over traditional computer implementations in terms of processing speed, evaluation accuracy, data management consistency, and real-time decision-making capabilities.

[0452] use Figure 14 The processing procedure is explained.

[0453] Step 1: Users input business topics and attribute information using the terminal.

[0454] Users open the application page on the terminal, enter the department's business theme (e.g., "reduce production defect rate") and their own attribute information (e.g., "Operator Level A" "Production Line Group 1") in the input box, and select parameters such as evaluation cycle from the drop-down menu or checkbox.

[0455] Input: Business topic text, user attribute information, evaluation period information.

[0456] The terminal packages the above input into structured data (e.g., key-value pairs) and sends it to the server via a secure communication protocol.

[0457] Output: Initial configuration data packets sent from the terminal to the server.

[0458] Step 2: The server generates a prompt statement for setting the target based on the business theme and attribute information.

[0459] After receiving the initial configuration data sent by the terminal, the server parses out the business theme and user attribute fields. Internally, the server calls the template engine to embed these fields into a predefined prompt statement template, generating natural language text describing the target task.

[0460] Input: Business topic, user attribute information, historical performance data (read by the server from the database).

[0461] The server performs statistical operations on historical performance data (such as calculating the average achievement and fluctuation range over the past few periods) and inserts the statistical results into prompts to increase the amount of contextual information in the generative artificial intelligence model.

[0462] Output: Target setting prompts containing business topics, user levels, and historical statistics.

[0463] Step 3: The server invokes a generative artificial intelligence model to generate tiered business objectives.

[0464] The server inputs the prompts generated in step 2 into the generative artificial intelligence model. The server encodes the prompts into a labeled sequence, performs forward propagation through the model's embedding layer and self-attention layer, calculates the hidden vector at each position, and finally generates the target description text in the output layer.

[0465] Input: Prompt statements for setting targets.

[0466] The server performs text parsing on the corresponding output text, extracting numbers, comparison symbols (such as "≤" and "≥"), and units (such as "%" and "times / day") into structured business objectives, such as "defect rate ≤ 1%" and "daily inspection ≥ 3 times".

[0467] Output: Target description text and structured hierarchical business target parameters.

[0468] Step 4: The server returns the business objective to the terminal, which then displays it.

[0469] The server packages the target description text and structured target parameters into response data and sends it to the terminal over the network.

[0470] Input: Target setting results (text + structured parameters).

[0471] After receiving the response, the terminal parses the JSON structure and calls the graphical control to display content such as "This cycle target: defect rate ≤1%, daily inspection ≥3 times" on the interface.

[0472] Output: Visualized target information displayed on the terminal screen.

[0473] Step 5: At the end of the cycle, users input the content and quantity information through the terminal.

[0474] On the terminal's "Results Reporting" page, users enter quantitative information such as the actual output, number of defective items, and number of completed tasks for the current cycle, and input descriptive text (such as "The main bottleneck is in the packaging process") in the text area.

[0475] Input: Quantitative data such as actual output and number of defective items, and explanatory text.

[0476] The terminal assembles the input data into a structured message and uploads it to the server via a security protocol.

[0477] Output: The data packets containing the completed content sent from the terminal to the server.

[0478] Step 6: The server calculates the numerical achievement once and generates features for evaluation.

[0479] After receiving the achievement message, the server reads the target parameters of the corresponding user from the database (e.g., target output of 500 pieces, defect rate ≤1%). The server performs arithmetic operations: calculates the actual defect rate = number of defective pieces / actual output, calculates the output achievement ratio = actual output / target output, and normalizes these values ​​(subtract the mean and divide by the standard deviation).

[0480] Input: Actual quantity data, target parameters.

[0481] The server constructs a feature vector, including production achievement ratio, defect rate deviation, historical average achievement, etc., and inputs the features into a small feedforward neural network to perform matrix multiplication and activation function operations to obtain a numerical achievement (e.g., 0.85).

[0482] Output: A numerical achievement and an evaluation feature vector containing multiple dimensions.

[0483] Step 7: The server generates a judgment and improvement suggestion using prompt statements and calls the generative artificial intelligence model.

[0484] Based on the numerical achievement level, target parameters, actual results, and user-descriptive text obtained in step 6, the server generates a prompt statement for initial assessment and improvement suggestions. For example: "Target: Daily output 500 units, defect rate ≤1%. Actual: Output 450 units, defective units 4, defect rate approximately 0.89%. Please evaluate the achievement in Simplified Chinese and provide three specific improvement suggestions that can be implemented within one week." Inputs: target parameters, actual data, first-time numerical achievement rate, and user-defined explanatory text.

[0485] The server encodes the prompt statement and inputs it into a generative artificial intelligence model. The model performs forward computation and outputs text containing evaluation and suggestion paragraphs.

[0486] Output: Achievement assessment text and business improvement suggestion text.

[0487] Step 8: The terminal collects the user's facial expressions and voice, and the server performs emotion analysis.

[0488] When users view the evaluation results, the terminal uses the camera to capture the user's facial image after obtaining the user's authorization, and records the user's brief voice feedback through the microphone.

[0489] Input: Facial image frames, audio signals.

[0490] The terminal performs basic compression and format conversion on the images, encodes the speech, and sends it to the server. The server performs face detection and cropping on the images, inputs the standardized face images into a convolutional neural network, and calculates the probability distribution of each emotion category. The server extracts acoustic features from the speech data and inputs them into a speech emotion recognition model to obtain the speech emotion category probability. The server fuses the two probabilities using either a weighted average or maximum a posteriori rule to determine the final emotion state label.

[0491] Output: Affective state labels and their confidence levels for subsequent evaluation and correction.

[0492] Step 9: The server determines whether to perform a second judgment based on the first judgment and the emotional state, and generates a prompt statement for the second judgment.

[0493] The server retrieves the numerical achievement rate and evaluation text from the database, and reads the sentiment status label generated in step 8. The server then performs a conditional judgment: when the numerical achievement rate is below a preset threshold and the sentiment status is negative (e.g., "frustrated"), the server reads previously implemented improvement measures from the historical records. It combines the initial judgment, sentiment status, and historical improvement implementation information to generate a secondary judgment prompt, such as: "The user's achievement rate this period is 60%, and the current sentiment status is 'frustrated.' Improvement measures taken in the previous period included 'adjusting the material replenishment frequency.' Please provide more encouraging evaluation text while maintaining the authenticity of the evaluation, and set two moderately difficult but quantifiable goals with room for improvement for the next period." Input: One judgment result, emotional state, and historical improvement implementation information.

[0494] The server inputs the prompt into the generative artificial intelligence model, performs inference operations to obtain the adjusted evaluation text and the target plan for the next cycle.

[0495] Output: Secondary evaluation text and target plan for the next cycle.

[0496] Step 10: The server update final achievement assessment and visualization data were generated.

[0497] Based on the results of the secondary judgment, the server may make smooth adjustments to the achievement of the primary score, such as applying a smoothing function to reduce excessive penalties in the low score range. The server then writes the updated final achievement to the database.

[0498] Input: First-order numerical achievement rate, second-order judgment information.

[0499] The server collects historical achievement rates, current final achievement rates, and real-time achievement rates of each process on the timeline in the visualization data generation module, constructs a chart data structure (including time arrays, value arrays, color configurations, etc.), and sets the color scheme and emphasis level according to the emotional state, encoding the achievement rate and emotional state together into the chart configuration.

[0500] Output: Visualization configuration data containing chart data and display style information.

[0501] Step 11: The terminal receives visual configuration data and presents evaluation results and suggestions.

[0502] After the terminal obtains the visualization configuration data from the server, it parses the time series, achievement value, and color and line type parameters, and calls the graphics drawing component to generate line charts, bar charts, or dashboards on the screen. At the same time, it displays the final achievement evaluation text, business improvement suggestions, and the target plan for the next cycle in the text area.

[0503] Inputs: Visualized configuration data, evaluation text, improvement suggestions, and target solutions for the next cycle.

[0504] The terminal adjusts the interface theme color based on style parameters related to emotional state. For example, it uses bright colors when the emotion is positive and softer dark colors when the emotion is negative, in order to enhance the user's understanding and acceptance of the system feedback.

[0505] Output: A comprehensive evaluation interface presented on the terminal display screen in the form of charts and text.

[0506] Step 12: The server uses real-time data and generative artificial intelligence models to generate process improvement suggestions in production process scenarios.

[0507] The server continuously receives status and performance information for each process from the information processing unit of the production line, stores this data in a time-series database with timestamps as keys, and performs sliding window statistical calculations for each process, such as calculating the average output, defect rate, and cycle time deviation over the most recent 30 minutes.

[0508] Inputs: Real-time status information, performance information, and process target parameters.

[0509] The server compares the window statistics with the process targets, calculates the real-time achievement rate of process units, and when the achievement rate of multiple consecutive windows falls below the benchmark value, the server generates a prompt statement containing the process name, current achievement rate, cycle time deviation, and recent trend, and inputs it into the generative artificial intelligence model to request process improvement suggestions. The server performs semantic analysis on the text output by the model, extracts key actions (such as "increase staff", "adjust material replenishment frequency", "optimize workstation layout") into structured suggestions, and pushes these suggestions to the workshop terminal or control system.

[0510] Output: Structured process improvement suggestions for specific procedures and corresponding explanatory text.

[0511] 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.

[0512] 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.

[0513] 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.

[0514] 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.

[0515] 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.

[0516] Second Implementation Method Figure 3 An example of the configuration of the data processing system 210 according to the second embodiment is shown.

[0517] 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.

[0518] 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).

[0519] 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.

[0520] 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.

[0521] 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).

[0522] 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.

[0523] 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.

[0524] 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.

[0525] 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).

[0526] 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.

[0527] 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".

[0528] 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.

[0529] 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.

[0530] 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.

[0531] 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.

[0532] 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.

[0533] 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.

[0534] 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.

[0535] 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.

[0536] 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.

[0537] Third Implementation Method Figure 5 An example of the configuration of the data processing system 310 according to the third embodiment is shown.

[0538] 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.

[0539] 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).

[0540] 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.

[0541] 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.

[0542] 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).

[0543] 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.

[0544] 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.

[0545] 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.

[0546] 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.

[0547] 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.

[0548] 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".

[0549] 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.

[0550] 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.

[0551] 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.

[0552] 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.

[0553] 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.

[0554] 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.

[0555] 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.

[0556] 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.

[0557] 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.

[0558] Fourth Implementation Method Figure 7 An example of the configuration of the data processing system 410 according to the fourth embodiment is shown.

[0559] 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.

[0560] 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).

[0561] 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.

[0562] 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.

[0563] 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).

[0564] 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.

[0565] 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.

[0566] 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.

[0567] 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.

[0568] 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.

[0569] 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.

[0570] 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".

[0571] 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.

[0572] 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.

[0573] 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.

[0574] 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.

[0575] 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.

[0576] 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.

[0577] 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.

[0578] 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.

[0579] 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.

[0580] 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.

[0581] 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.

[0582] 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.

[0583] 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).

[0584] 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.

[0585] 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."

[0586] 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.

[0587] 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).

[0588] 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.

[0589] 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.

[0590] 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.

[0591] 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.

[0592] 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.

[0593] 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.

[0594] 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.

[0595] 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.

[0596] 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.

[0597] 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.

[0598] In addition, the following notes are provided in response to the above explanation.

[0599] Example 1 (Note 1) An information processing system, characterized in that it comprises: An apparatus for receiving topic information and hierarchical information as conditions for setting organizational goals, receiving the topic information and hierarchical information as input information from a terminal, validating the content of the input information, converting the input information into a standardized internal representation, and storing it in association with a storage device. An apparatus for generating structured target information including period information and achievement indicators by performing calculations on the topic information and the hierarchical information based on a trained model trained using a machine learning algorithm, using historical information about past goals and achievement stored in the storage device, to calculate target parameters, and applying threshold conditions and constraints to the target parameters for correction. An apparatus for automatically generating prompt statements for inputting into a generative artificial intelligence model based on the structured target information, the topic information, and the hierarchical information; inputting the prompt statements into the generative artificial intelligence model to generate target text in natural language; and performing inappropriate expression detection and judgment processing on the generated target text. The device is used to output the target text to the terminal for presentation, obtain evaluation information or feedback information as a correction request from the terminal, store the feedback information in association with the historical information, and accumulate it as learning data for updating the trained model and prompt statement generation processing. An apparatus for generating a prompt statement for inputting into a generative artificial intelligence model based on actual information as input for the content to be achieved, obtaining a first-degree achievement result using the generative artificial intelligence model based on the prompt statement, inferring the emotional state from the user's voice and facial expression information using an emotion analysis algorithm, and correcting the first-degree achievement result based on the emotional state, thereby generating a degree of achievement assessment result. A device for generating display data that displays the achievement determination results and the target information in a graphical or chart-like manner on a visualization information display device, and presents them in a form that is easy for users to understand at a glance.

[0600] (Note 2) The information processing system according to Appendix 1 is characterized in that, It also includes: a device for generating a prompt statement for inputting into a generative artificial intelligence model based on the first determination result of the achievement level and the emotional state, obtaining a second determination result of the achievement level using the generative artificial intelligence model based on the second determination prompt statement, and reflecting the second determination result as the achievement level determination result in the displayed data.

[0601] (Note 3) The information processing system according to Appendix 1 is characterized in that, It also includes: a device for comprehensively analyzing the feedback information and the achievement determination result in conjunction with the historical information, thereby automatically updating the learning data of the trained model and the generation rules of the prompt statement, and performing model update processing in a periodic manner or under predetermined conditions, thereby improving the objectivity and consistency of goal setting and achievement determination in the organization.

[0602] Application Example 1 (Note 1) An information processing system, characterized in that it comprises: This is used to acquire business unit setting theme information and business difficulty differentiation information obtained by the computing unit in the information processing device, and based on the business unit setting theme information and the business difficulty differentiation information, to acquire a historical data set containing historical business performance data and quality-related data in order to set the corresponding target value, and to convert the historical data set into tabular data using data processing program and statistical learning program, to perform data preprocessing on the tabular data including feature extraction, missing value handling and indicator calculation, and to construct a regression operation model based on the preprocessed data, and to input the feature quantities representing the business unit setting theme and the business difficulty differentiation into the regression operation model to predict the target value; Based on the target value obtained by the target value prediction unit and the historical data set, the target description generation unit automatically generates a prompt statement for input to the generative artificial intelligence model, inputs the prompt statement into the generative artificial intelligence model to generate explanatory text including the reason for setting the target value and the running instructions corresponding to the target value, and records the explanatory text into the management storage unit and outputs it as information for display. This is used to acquire business achievement content data input by the user through a terminal device, compare the business achievement content data with the target value to calculate the achievement index, generate a prompt statement for input to a generative artificial intelligence model based on an evaluation data set containing the achievement index and the business achievement content data, input the prompt statement into the generative artificial intelligence model to generate a primary evaluation text related to the achievement, and determine a primary judgment unit for a primary judgment result of the achievement based on the primary evaluation text. A supplementary determination unit is used to perform emotion recognition processing on acquired voice and image information using an emotion analysis program, calculate an emotion index representing the user's emotional state, and correct the achievement determination based on the emotion index and the initial determination result to obtain a supplementary achievement determination result.

[0603] (Note 2) The information processing system according to Appendix 1 is characterized in that it further includes: a secondary judgment unit for generating a prompt statement for secondary evaluation input to a generative artificial intelligence model based on the primary judgment result obtained by the primary judgment unit, the corrected achievement information obtained by the supplementary judgment unit, and the past judgment history included in the historical data set; inputting the prompt statement for secondary evaluation into the generative artificial intelligence model to generate secondary evaluation text and a final achievement judgment; and outputting the final achievement judgment as the achievement evaluation result.

[0604] (Note 3) The information processing system according to Appendix 1 is characterized in that it further includes: a visualization data group for generating, based on the target value obtained by the target value prediction unit, the achievement evaluation results obtained by the primary judgment unit and the secondary judgment unit, and the time change information obtained based on the historical data set, a visualization display unit for generating, based on the target value obtained by the target value prediction unit, the achievement evaluation results obtained by the primary judgment unit and the secondary judgment unit, and the time change information obtained by the primary judgment unit, a visualization display unit for generating, based on the visualization data group, a display control signal for representing, using at least one graphic element, the relationship between the target value and the actual value, the achievement evaluation results, and the quality-related indicator changes, and displaying, based on the display control signal, the achievement evaluation results in the form of a chart on an information display device, so that individuals can clearly grasp the target achievement status at a glance.

[0605] Example 2 (Note 1) An information processing system, characterized in that it comprises: Executed by an information processing device, this unit generates prompt statements for setting multi-level targets based on the set theme of the business unit, and generates target information corresponding to multiple levels based on the prompt statements. Performed by an information input device, it is used to receive the content achieved by an individual through an interactive interface, and to acquire the content as text information and convert it into structured data. A unit executed by a relational data storage device for recording the structured data as data records and storing the achieved content in association with identification information and time information; This unit, executed by an information processing device, is used to extract keywords and perform statistical processing on multiple achieved contents stored in the relational data storage device using data analysis software, and to quantify the degree of achievement according to a predefined evaluation benchmark to generate a judgment result. A unit executed by an information processing device, which generates a prompt statement input to a generative artificial intelligence model based on the achieved content and the first judgment result, provides the prompt statement to the generative artificial intelligence model to obtain evaluation information for the achieved content, and integrates the first judgment result with the evaluation information to generate a comprehensive judgment result of the degree of achievement. A unit executed by an information processing device, used to analyze a user's voice and / or facial expressions using an emotion analysis engine to identify the user's emotions, and to correct the comprehensive judgment result based on the analysis results of the emotions; This unit, executed by an information display device, is used to visualize the comprehensive judgment result and its changes over time in the form of graphics or charts, thereby generating feedback information that individuals can intuitively understand.

[0606] (Note 2) The information processing system according to Appendix 1 is characterized in that, The information processing device is configured to: generate a prompt statement for secondary evaluation containing additional verification conditions based on the primary judgment result and the evaluation information obtained from the generative artificial intelligence model; input the prompt statement for secondary evaluation into the generative artificial intelligence model to obtain a secondary judgment result; and update the comprehensive judgment result of the achievement level based on the secondary judgment result.

[0607] (Note 3) The information processing system according to Appendix 1 is characterized in that, The information processing device is configured to: summarize the achievement content, quantified achievement degree and evaluation information generated by the generative artificial intelligence model for multiple periods stored in the relational data storage device, generate visualized data for time series comparison, hierarchical comparison and / or group comparison, and display the visualized data through the information display device.

[0608] Application Example 2 (Note 1) An information processing system, characterized in that it comprises: A processing device for generating prompt statements for inputting into a generative artificial intelligence model based on the organization's business themes and the attribute information of the users, and generating hierarchical business objectives based on the prompt statements and related historical information. A processing device for obtaining input business achievement content and quantity information from an information terminal at the end of a cycle, performing achievement degree numerical evaluation calculation based on the comparison result of the business achievement content and the business target, generating a prompt statement containing the comparison result and the achievement content, and using the prompt statement as input to generate an achievement degree judgment result and business improvement suggestions by a generative artificial intelligence model. This processing device is used to acquire facial expression information and voice information of the user from an image acquisition device and a voice acquisition device, input the facial expression information and voice information into an emotion analysis function to determine the emotional state of the user, and correct the achievement assessment based on the emotional state and the first determination result. At the same time, it generates a prompt statement to determine the display method reflecting the emotional state, and generates the achievement assessment text and the next cycle target plan by a generative artificial intelligence model. A processing device for storing the achievement assessment, the business improvement suggestions, and the next cycle target plan into an information storage device, and generating visualization data as display control information, the visualization data including display mode information corresponding to the achievement assessment and the emotional state, and outputting the visualization data to the information display device; This is used to obtain status and performance information from the running object process, statistically analyze the status and performance information within a predetermined time window, calculate the real-time achievement rate of the process unit based on the comparison with the business objective, generate a prompt statement containing the status information and the real-time achievement rate when the real-time achievement rate is lower than the benchmark value, generate process improvement suggestions by a generative artificial intelligence model, and provide the process improvement suggestions to the processing device of the information display device or control device.

[0609] (Note 2) The information processing system according to Appendix 1 is characterized in that, When the primary judgment result and the emotional state meet predetermined conditions, the processing device generates a prompt statement for secondary judgment, which includes the primary judgment result, the emotional state, and historical business improvement implementation information. The prompt statement for secondary judgment is used as input to generate a secondary judgment result of achievement and a next cycle target adjusted for difficulty by a generative artificial intelligence model. The final achievement assessment is updated based on the secondary judgment result.

[0610] (Note 3) The information processing system according to Appendix 1 is characterized in that, Based on the visualized data, the processing device integrates the achievement assessment and the achievement divided by process unit as indicators along the time axis, and generates a graphic or chart that changes color scheme and emphasis according to the emotional state and achievement level. This graphic or chart is then displayed by the information display device in a way that is easy for the user to understand at a glance. The process improvement suggestions and the next cycle target plan are also displayed in association with the graphic or chart.

Claims

1. An information processing system, characterized in that, include: processor; The processor is configured as follows: Based on the department's set theme, a generative artificial intelligence model is used to generate prompts for setting goals for each job level, and corresponding goals are generated based on these prompts. Based on the input content, a generative artificial intelligence model is used to generate prompts for determining the degree of achievement, and the degree of achievement is determined based on the prompts. To identify user emotions, an emotion analysis engine is used to analyze the user's voice and facial expressions, and the determination of achievement is supplemented based on the identified emotions.

2. The information processing system according to claim 1, characterized in that, The processor is further configured to: generate a prompt message for a second determination of the achievement degree based on the first determination result of the achievement degree using a generative artificial intelligence model, and perform the second determination of the achievement degree according to the prompt message.

3. The information processing system according to claim 1, characterized in that, The processor is also configured to: use an information display device to display the achievement determination result in the form of a graph or chart, so that an individual can understand the achievement determination result at a glance.

Images