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 link a personnel management device, a performance evaluation device, a target setting device, a business progress analysis device, and a machine learning device to analyze the evaluation, target, and business progress of employees; the processor is configured to generate prompt information for instructing a generative artificial intelligence model to perform a specific analysis or prediction; the processor is configured to cause the generative artificial intelligence model to perform analysis or prediction based on the prompt information, and visualize the results of the analysis or prediction.

Description

Technical Field

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

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

[0003] The existing personnel management, performance evaluation, goal setting, and business progress analysis are mostly performed by different systems or devices, lacking effective linkage between them. This leads to the following problems: First, it is difficult to unify and integrate employees' historical evaluation data, current goals, and real-time business progress information, making it impossible for managers and employees to grasp employees' overall performance and goal achievement from a holistic perspective. Second, traditional evaluation and analysis methods rely heavily on manually setting rules or manually summarizing data, making it difficult to conduct timely and dynamic in-depth analysis or prediction, especially in providing personalized improvement suggestions based on individual business characteristics. Third, existing systems focus more on outcome-based evaluation, lacking a mechanism for relative comparison based on the progress of other members in the same department or team. Employees find it difficult to understand their relative position in the team, and managers find it difficult to identify those who need focused guidance in a timely manner. Fourth, evaluations are mostly concentrated at formal assessment points such as annual or quarterly assessments, lacking frequent feedback on potential performance evaluations, making it difficult to continuously motivate employees in daily work and support managers in conducting refined and continuous management of subordinates' growth.

[0004] Therefore, there is a need for a system that can link personnel management devices, performance evaluation devices, target setting devices, business progress analysis devices, and machine learning devices, and use generative artificial intelligence models to automatically generate prompts to perform specific analyses or predictions. This would enable data integration, intelligent analysis, result visualization, and feedback on relative and periodic suspected evaluations on a unified platform, thereby solving the aforementioned problems. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides an information processing system. The system includes a processor configured to link personnel management devices, performance evaluation devices, target setting devices, business progress analysis devices, and machine learning devices to comprehensively analyze employee evaluations, targets, and business progress. By collecting, correlating, and fusing data from different devices under unified processor control, a multi-dimensional data foundation covering employee historical performance, current targets, and real-time progress can be constructed, providing a reliable data source for subsequent intelligent analysis and prediction.

[0006] The processor is further configured to generate prompts that instruct the generative AI model to perform specific analyses or predictions. Specifically, based on employee-related data acquired from the aforementioned devices and a predetermined analytical objective, the processor automatically constructs prompts conforming to the input format of the generative AI model. These prompts trigger the generative AI model to analyze or predict aspects such as employee goal achievement, business behavior patterns, or future trends. In this way, the natural language understanding and generation capabilities of the generative AI model can be leveraged to achieve flexible descriptions and multi-dimensional reasoning of complex business scenarios, avoiding the difficulties traditional rule-based systems face in handling unstructured information or complex relationships.

[0007] The processor is also configured to enable the generative artificial intelligence model to perform analysis or prediction based on the prompt information, and to visualize the results of the analysis or prediction. The processor receives the analysis or prediction results output by the generative artificial intelligence model, and transforms them into structured data and visualizations suitable for human-computer interface display, such as charts, indicator overviews, and text descriptions, so that employees and managers can intuitively understand the current degree of employee goal achievement, key issues, and potential risks, and make it easier to formulate specific improvement actions accordingly.

[0008] To further address the challenges of employees lacking relative positional references and managers struggling to promptly identify problematic individuals, the processor is also configured to generate prompts for calculating relative probable performance evaluations based on the work progress of members in other departments. The generative AI model then uses these prompts to calculate the relative probable evaluations. By incorporating progress data from other members of the same department or team into the prompts, the generative AI model can output indicators reflecting an employee's relative performance within the group, such as a relative probable score, percentile rank, or ranking range. This relative probable evaluation allows employees to objectively understand their position within the team, while managers can quickly identify high-performing employees or those requiring focused guidance, thereby improving overall team management efficiency.

[0009] To provide continuous feedback and enhance employee motivation in daily work, the processor is configured to periodically generate prompts, calculate suspected evaluations based on these prompts, and provide these suspected evaluations to users to increase user motivation and facilitate managers' management of subordinates' growth. Specifically, the processor can automatically trigger the generation of prompts on a monthly, weekly, or other predetermined basis, analyze employees' goal achievement and business progress within that period, calculate suspected evaluations, and provide suspected evaluation results and related suggestions to employees and managers through a user interface or notification mechanism. Employees can thus receive suspected feedback close to formal evaluation results at a higher frequency, allowing them to adjust their behavior promptly; managers can dynamically grasp subordinates' growth trajectories and status changes based on continuously updated suspected evaluation data, implementing more targeted guidance and resource allocation. Through these methods, the present invention effectively solves the problems of scattered data, static analysis, lack of relative comparison and high-frequency feedback in existing technologies, achieving intelligent, visualized, and dynamic management of employee performance and growth processes.

[0010] "System" refers to an entire system consisting of hardware and / or software, including at least one processor and personnel management devices, performance evaluation devices, target setting devices, business progress analysis devices and machine learning devices that can communicate with the processor, for performing the data linkage, analysis and evaluation processing described in this invention.

[0011] A processor is a hardware unit or combination thereof capable of executing computer-readable instructions to perform functions such as data acquisition, data processing, prompting information generation, calling generative artificial intelligence models, and visualizing results. Examples include a central processing unit (CPU), a graphics processing unit (GPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or any combination thereof.

[0012] "Personnel management device" refers to a device or system used to manage human resources-related data such as basic employee information, personnel files, organizational structure, employment relationships, and attendance records. It can be a standalone server, a personnel management software platform, or a combination thereof.

[0013] "Performance evaluation device" refers to a device or system used to record, calculate and store employee performance scores, evaluation indicators and evaluation results within a specific period. It can generate formal performance evaluation data according to predetermined evaluation rules, indicator system and scoring model.

[0014] "Goal setting device" refers to a device or system used to set, record and manage individual or team business goals, performance goals and their weights, deadlines and measurement methods for employees, and to provide employees with target benchmark data for each evaluation period.

[0015] "Business progress analysis device" refers to a device or system used to collect and analyze progress data generated by employees in the process of performing specific business tasks, including but not limited to task completion status, sales process data, project milestone completion status, customer visit records, etc., and supports the statistical analysis of this data.

[0016] "Machine learning device" refers to a device or system used to perform machine learning algorithms, model training, model updating and inference calculations. It may include machine learning frameworks, model service platforms and corresponding computing hardware resources, and is used to model and extract features from employee-related data.

[0017] "Generative AI models" refer to AI models that can generate text, structured data, or other forms of output based on input prompts, including but not limited to large language models, text generation models, or multimodal models with generative capabilities, used to perform specific analysis or prediction tasks based on prompts.

[0018] "Prompt information" refers to the input content of a generative artificial intelligence model, which is automatically constructed by the processor according to a predetermined purpose and available data. It includes natural language text, structured parameters or combinations thereof, and is used to instruct the generative artificial intelligence model to perform corresponding analysis, prediction or evaluation calculations.

[0019] "Analysis or prediction" refers to the processing of information and relevant data by generative artificial intelligence models, including current situation analysis, trend prediction, risk assessment, goal achievement assessment, problem identification, and generation of improvement suggestions.

[0020] "Visualization" refers to the process of converting analysis or forecast results into graphical, chart-based, or structured presentations so that users can intuitively understand the results on the interface, such as displaying them in the form of line charts, bar charts, pie charts, indicator panels, or text summaries.

[0021] "Relative suspected evaluation" refers to the evaluation result calculated based on the comparison of the target employee's business progress data with other department members in the same period, which is used to reflect the employee's relative performance in the group. This evaluation does not have to be a formal assessment result, but rather a relative score, ranking range or percentile indicator inferred by a generative artificial intelligence model.

[0022] "Suspected evaluation" refers to an estimating evaluation that approximates the formal evaluation result, which is derived by the system based on the employee's goals, performance and business progress data within a specific period at informal assessment points and analyzed by a generative artificial intelligence model. It is used to provide a reference for employees and managers in daily work.

[0023] "User" refers to the subject that interacts with the system of this invention through a terminal, including but not limited to the employee being evaluated, the manager who manages subordinates, the human resources manager, or other users with corresponding access permissions. Attached Figure Description

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

[0025] Figure 2 This is a conceptual diagram illustrating an example of the main functions of the data processing apparatus and smart device according to the first embodiment.

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

[0027] Figure 4 This is a conceptual diagram illustrating an example of the main functions of the data processing device and smart glasses according to the second embodiment.

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

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

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

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

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

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

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

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

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

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

[0038] Hereinafter, an example of an implementation of the system to which the technology of this disclosure relates will be described with reference to the accompanying drawings.

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

[0040] In the following embodiments, the processor (hereinafter referred to as "processor") with reference numerals may be a single computing device or a combination of multiple computing devices. Furthermore, the processor may be a single computing device or a combination of multiple computing devices. Examples of computing devices include CPU (Central Processing Unit), GPU (Graphics Processing Unit), GPGPU (General-Purpose Computing on Graphics Processing Units), APU (Accelerated Processing Unit), etc.

[0041] In the following embodiments, RAM (Random Access Memory), as indicated in the figures, is a memory that temporarily stores information and is used as working memory by the processor.

[0042] In the following embodiments, the memory, as indicated by the reference numerals, is one or more non-volatile storage devices that store various programs and parameters. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), disks (e.g., hard disks), or magnetic tapes.

[0043] In the following embodiments, the communication I / F (interface) with reference numerals is an interface that includes a communication processor and an antenna, etc. The communication I / F is responsible for communication between multiple computers. As an example of a communication specification applicable to the communication I / F, wireless communication specifications such as 5G (5th Generation Mobile Communication System), Wi-Fi (wireless fidelity) (registered trademark), or Bluetooth (registered trademark) can be listed.

[0044] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B". That is, "A and / or B" means that it can be only A, only B, or a combination of A and B. Furthermore, in this specification, when "and / or" connects to express more than three items, the same interpretation as "A and / or B" applies.

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

[0046] like Figure 1 As shown, the data processing system 10 includes a data processing device 12 and an intelligent device 14. A server can be cited as an example of the data processing device 12.

[0047] The data processing apparatus 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" as understood in this disclosure. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0048] The smart device 14 includes a computer 36, a receiving device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. In addition, the receiving device 38, output device 40, camera 42, and communication I / F 44 are also connected to the bus 52.

[0049] The receiving device 38 includes a touchscreen 38A and a microphone 38B, and receives user input. The touchscreen 38A receives user input via touch by detecting contact with an indicator (e.g., a pen or finger). The microphone 38B receives user input via sound by detecting the user's voice. The control unit 46A in the processor 46 sends data representing the user input received by the touchscreen 38A and microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data representing the user input.

[0050] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting data in a form perceptible to the user 20 (e.g., sound and / or text). The display 40A displays visual information such as text and images according to instructions from the processor 46. The speaker 40B outputs sound according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge-Coupled Device) image sensor.

[0051] Communication I / F44 is connected to network 54. Communication I / F44 and 26 are responsible for sending and receiving various information between processor 46 and processor 28 via network 54.

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

[0053] like Figure 2 As shown, in the data processing apparatus 12, specific processing is performed by the processor 28. A specific processing program 56 is stored in the memory 32. The specific processing program 56 is an example of a "program" as understood in this disclosure. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0054] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. In the emotion inference function (emotion-specific function) using the emotion-specific model 59, various inferences and predictions related to the user's emotions are performed, including inferences and predictions of the user's emotions, but this is not limited to this example. Furthermore, emotion inference and prediction may also include, for example, emotion analysis (parsing).

[0055] In the smart device 14, the processor 46 performs the acceptance output processing. The memory 50 stores the acceptance output program 60. The acceptance output program 60 is used in conjunction with the data processing system 10 and the specific processing program 56. The processor 46 reads the acceptance output program 60 from the memory 50 and executes the read acceptance output program 60 on the RAM 48. The acceptance output processing is implemented by the processor 46 acting as the control unit 46A according to the acceptance output program 60 executed on the RAM 48. Furthermore, the smart device 14 has the same data generation model and emotion-specific model as the data generation model 58 and the emotion-specific model 59, and these models can also be used to perform the same processing as the specific processing unit 290. The acceptance output processing is implemented by the processor 46 acting as the control unit 46A according to the acceptance output program 60 executed on the RAM 48.

[0056] Alternatively, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device (e.g., a generation server) may have the data generation model 58. In this case, the data processing device 12 obtains the processing results (prediction results, etc.) using the data generation model 58 by communicating with the server device that has the data generation model 58. Furthermore, the data processing device 12 may be a server device or a user-held terminal device (e.g., a mobile phone, robot, home appliance, etc.). Next, an example of the processing of the data processing system 10 of the first embodiment will be described.

[0057] Example 1 The flow of a specific process in Example 1 will be described. Each part of the system described below is implemented by the data processing device 12 and the smart device 14. Furthermore, the data processing device 12 is referred to as the "server," and the smart device 14 is referred to as the "terminal."

[0058] With the widespread adoption of systems for human resource management, performance evaluation, goal management, and business progress management, enterprises have generated a large amount of structured and semi-structured data. Current technologies typically perform simple statistical analysis or fixed-rule performance calculations within a single business system, which presents the following technical problems: (1) Data between various business systems is stored in isolation, lacking a unified data integration and feature processing mechanism, which makes it difficult to form high-quality datasets that can be directly used for prediction and intelligent analysis in the computer, thus limiting the data-driven decision support capability.

[0059] (2) Traditional analysis engines based on fixed algorithms or predefined rules cannot flexibly construct complex query and analysis logic according to the specific contextual needs of users when running on the server side, making it difficult for the same set of data to respond to the diverse analysis requests of managers at different levels and individual employees in a timely manner.

[0060] (3) Even if machine learning models are introduced to predict the probability of achieving goals or efficiency indicators, existing systems usually implement machine learning modules separately from other business modules. They do not form an end-to-end computing process from data collection, feature generation, model training, prediction inference to result interpretation within the server. Therefore, it is difficult to update the model input features and output interpretation in a timely manner and cannot continuously reflect the dynamic changes in organizational and individual behavior.

[0061] (4) The application of existing generative artificial intelligence models is mostly limited to open-ended question answering or general text generation. They lack a tight coupling mechanism with the enterprise's internal structured data, machine learning prediction results and visualization results. The server cannot automatically construct prompts for specific users and specific business scenarios, making it difficult to generate highly executable personalized suggestions while ensuring data consistency.

[0062] (5) In terms of relative evaluation and continuous incentive, traditional systems usually only provide static rankings or one-time performance reports. They lack dynamic relative evaluation calculation mechanisms based on group distribution and time series, and also lack calculation processes that periodically trigger analysis, generate feedback and push it to user terminals on the server side. This is not conducive to achieving the technical effect of continuous guidance of user behavior and growth path planning through computer means.

[0063] Therefore, this invention aims to construct a unified processing architecture that integrates data integration, feature generation, predictive model training and inference, visualization generation, and generative artificial intelligence model invocation within a server, from a computer technology perspective. By automatically generating and inputting prompts, the server can efficiently produce interpretive analysis results and actionable suggestions for individuals and groups while ensuring data security and consistency. This will improve the computer's data processing capabilities, intelligent analysis capabilities, and human-computer interaction capabilities in performance management and goal management scenarios.

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

[0065] In this invention, the server includes: a device for acquiring structured data from a personnel information processing device, a business evaluation processing device, a target management processing device, a business progress analysis processing device, and a learning processing device via a communication interface, and integrating the structured data based on identifiers and storing it in a data storage device; a device for executing a data parsing program on the structured data stored in the data storage device to perform missing value completion, standardization, summarization, and feature generation, thereby generating a set of parsed data containing evaluation indicators, target achievement indicators, and business progress indicators for operators; a device for learning a prediction model using a prediction model building program based on the parsed data set, and using the prediction model to estimate the target achievement probability and business efficiency indicators of operators, storing the obtained estimation results as estimation result data in the data storage device; and a device for generating, using a visualization generation program based on the parsed data set and the estimation result data, information including time change information, achievement probability information, and group ranking. The apparatus comprises: a device for visualizing information data and outputting the visualized data as visual information that can be displayed on a display device; a device for obtaining, based on user identification information, the parsed data set and the inference result data related to the user from the data storage device, summarizing and extracting the obtained data to generate background information containing the user's business status, prediction results, and comparison information with others; a device for automatically generating prompt statements containing analysis request content, object scope, and output format conditions based on the background information and instruction rules for a generative artificial intelligence model; a device for combining the prompt statements and the background information to form input information for a generative artificial intelligence model, calling the generative artificial intelligence model, performing analysis processing or prediction processing according to the input information, and obtaining a generation result containing suggestion information and explanatory information; and a device for outputting the generation result in association with the visual information, and providing support information to the user for formulating or revising business plans based on the associated output. This enables an end-to-end computing process within the same server architecture, encompassing multi-source business data integration, feature generation, predictive inference, natural language interpretation, and suggestion output. By automatically generating and inputting prompts tightly coupled with background information, the output of generative artificial intelligence models is kept consistent with structured prediction and visualization results. This improves the data processing efficiency, prediction accuracy, and human-computer interaction quality of computers in performance and goal management scenarios, enabling dynamic relative evaluation and continuous incentives for users, and enhancing the overall technical performance of the information processing system.

[0066] "Personnel information processing device" refers to an electronic device or program module used to collect, store, and manage personnel-related data such as identity information, appointment information, job information, and organizational information in a computer system.

[0067] "Business evaluation processing device" refers to an electronic device or program module used to record, calculate and manage business performance, performance scores, key indicator completion and other evaluation-related data in a computer system.

[0068] "Target management processing device" refers to an electronic device or program module used in a computer system to set, update, and manage business objectives, weights, deadlines, and target status information.

[0069] "Business progress analysis and processing device" refers to an electronic device or program module used in a computer system to collect, statistically analyze, and process business progress-related data such as task progress, project status, completion rate, and time consumption records.

[0070] "Learning processing device" refers to an electronic device or program module used to perform data-driven learning computations in a computer system to build or update predictive models, classification models or other statistical models.

[0071] "Communication interface" refers to a hardware or software interface used to transmit data between different electronic devices, program modules or network nodes, including but not limited to network interfaces, application programming interfaces or message queue interfaces.

[0072] "Structured data" refers to data organized according to a predefined pattern or data structure and stored in the form of fields, including tabular data, record data, or serialized data with fixed fields.

[0073] An "identifier" is a data item that can uniquely or quasi-uniquely distinguish an object, entity, or record in a computer system, including but not limited to user ID, employee number, device ID, or record primary key.

[0074] "Data storage device" refers to a hardware device or software system used to store data records, indexes and metadata on non-volatile or volatile storage media, including database systems, file systems or key-value stores.

[0075] A "data parsing program" refers to a software program or script that runs on a processing device and is used to read, verify, convert formats, handle missing values, standardize, and summarize raw data.

[0076] "Missing value completion" refers to the process of filling in estimated or default values ​​in a computer with missing or blank fields in a dataset using rules or algorithms.

[0077] "Standardization" refers to the process of converting data with different dimensions or different value ranges into a unified scale or distribution form so that they can be compared, aggregated, or modeled in a computer.

[0078] "Summary" refers to the process of performing counting, summing, averaging, maximum value, minimum value or other aggregation operations on detailed data based on one or more grouping conditions.

[0079] "Feature generation" refers to the data processing process that generates new feature variables from the original data through calculation, combination, or transformation, so that they can be used as model inputs or for further analysis.

[0080] "Analyzed dataset" refers to a collection of cleaned, transformed, and expanded data records that have been processed after data parsing and feature generation, and is used for subsequent predictive model training and inference.

[0081] "Predictive model building program" refers to a software program or script that runs on a processing device and is used to train a statistical or machine learning model based on a parsed dataset and save the model parameters.

[0082] A "predictive model" is a mathematical or computational model that is formed by learning the relationship between inputs and outputs in historical data and can be used to predict or infer unknown samples.

[0083] "Probability of achieving the goal" refers to the probability index in the output of the prediction model, which represents the likelihood that a specific goal will be achieved within a predetermined time frame.

[0084] "Business efficiency indicators" refer to calculation indicators used to quantitatively describe the quantity, quality, or resource utilization of tasks completed by personnel within a unit of time.

[0085] "Inferred result data" refers to the output results obtained by inferring from the analytical data set through a prediction model, and the data set is stored in the data storage device in the form of records.

[0086] "Visualization generation program" refers to a software program or script that runs on a processing device and is used to convert numerical or categorical data into visual representations such as charts, graphs, or dashboards.

[0087] "Visualized data" refers to a collection of information generated by a visualization generation program that contains numerical points, coordinates, color codes, or chart configurations used to draw graphic elements.

[0088] "Visual information" refers to graphics, charts, or combined views presented on a display device, based on visualized data, used to intuitively show data characteristics and trends.

[0089] "Time change information" refers to sequence information used to represent how data changes over time, including timestamps, time intervals, and corresponding values ​​or states.

[0090] "Intragroup ranking information" refers to the ranking, percentile, or relative position of objects within a group, obtained by ranking them according to one or more indicators.

[0091] "User identification information" refers to data items used in the system to uniquely or quasi-uniquely identify a user, including account identifiers, login identifiers, or device identifiers.

[0092] "Background information" refers to a set of data or text information that is obtained from a data storage device based on user identification information and is processed through summarization and extraction to describe the business status, prediction results, and comparisons with others of a specific user.

[0093] "Generative artificial intelligence models" refer to artificial intelligence models that are trained on a large amount of sample data and can automatically generate natural language text or other forms of output based on input information.

[0094] "Instruction rules" refer to a set of rules used to define how to generate prompt statements from background information and system configuration, including definitions of analysis content, target objects, output formats, and constraints.

[0095] "Prompt statements" refer to one or more pieces of text automatically constructed according to instruction rules and used as input to a generative artificial intelligence model to instruct the model to perform specific analysis, prediction, or suggestion generation tasks.

[0096] "Input information" refers to the unified input content, which is composed of prompts and background information, and is fed into the generative artificial intelligence model for reasoning.

[0097] "Suggestion information" refers to the text content output by a generative artificial intelligence model after analyzing the input information, which contains operational suggestions, strategies, or optimization solutions for a specific object or situation.

[0098] "Explanatory information" refers to the explanatory text output by generative artificial intelligence models that explains prediction results, the meaning of indicators, or the basis for recommendations.

[0099] "Generated results" refers to the overall set of results output by a generative artificial intelligence model after receiving input information, including suggestions, explanations, and other relevant text or structured content.

[0100] "Group information processing program" refers to a program module or script that is executed on a processing device to perform statistical analysis, distribution calculation, and relative comparison of data from multiple members.

[0101] "Benchmark distribution information" refers to statistical information calculated based on population data to represent the distribution characteristics of one or more indicators within a population, including forms such as mean, variance, quantiles, or histograms.

[0102] "Relative evaluation information" refers to information that represents an individual's relative position or relative level in a group, calculated based on the relationship between baseline distribution information and individual indicator values.

[0103] "Suspected evaluation" refers to an evaluation result derived from relative evaluation information or other indirect indicators, which is used to approximately reflect an individual's performance level, and is not necessarily completely equivalent to a formal performance score.

[0104] "Improvement plan" refers to specific action suggestions or plans generated by the system or generative artificial intelligence model based on evaluation results or relative evaluation information, for the purpose of improving performance or achieving goals.

[0105] "Time-based control program" refers to a program module that executes on the processing device and periodically triggers operations such as data processing, model invocation, and result notification based on timestamps, periodic configurations, or scheduling rules.

[0106] "User terminal" refers to an electronic device operated by a user for interacting with a server via a network to receive visual information and generate results, including computer terminals, mobile terminals, or other devices with display and input functions.

[0107] In one embodiment of the invention, the server is a computing device deployed in a data center or enterprise intranet. The server includes a multi-core central processing unit, memory, non-volatile storage, and a network interface. The server runs a general-purpose operating system, such as a server operating system based on an open-source kernel, and installs a relational database management system, such as relational database software, a data analysis runtime environment, such as an interpreted programming language environment, and a visualization platform, such as a general-purpose business intelligence visualization platform. Furthermore, the server deploys a generative artificial intelligence model inference service, which provides an interface for text-based generative artificial intelligence models.

[0108] In a preferred embodiment of the present invention, the server loads multiple software modules into its memory. These modules logically include: a data acquisition module, a data preprocessing and feature generation module, a prediction model module, a visualization generation module, a background information generation module, a prompt statement generation module, a generative artificial intelligence model interface module, and a result integration and output module. The modules exchange data with each other through data structures in memory and data tables in a database, thereby forming an end-to-end computational data flow.

[0109] In terms of data storage structure, the server creates multiple tables in the relational database to store personnel information, evaluation records, target records, progress records, feature data, prediction result data, relative evaluation data, and generated result data. For example, the server can create tables of the following types in the database: Employee Basic Information Table (containing employee ID, organization ID, and job category fields), Performance Record Table (containing employee ID, time period, score, and sub-dimension score fields), Target Record Table (containing employee ID, target ID, weight, deadline, and target difficulty level), Progress Record Table (containing target ID, timestamp, current completion percentage, and actual time spent), Feature Data Table (containing employee ID, time period, and derived feature values), Prediction Result Table (containing employee ID, target ID, predicted achievement probability, and predicted efficiency score), Group Statistics Table (containing department ID, indicator name, mean, variance, and quantiles), and Generated Result Table (containing user ID, timestamp, original prompt statement, generated text result, and related predicted value references), etc.

[0110] In terms of data parsing and feature generation, the server uses data table structure objects from the data analysis software environment to load structured data collected from various business processing devices into memory. The server unifies data from different sources into a predefined internal format through field mapping and type conversion; for example, dates are unified as timestamp integers, scores as floating-point numbers, and category fields as enumerated codes. Based on this, the server executes missing value completion algorithms, such as using grouped averages or medians to fill in missing values ​​for continuous indicators, and mode to fill in missing values ​​for discrete indicators; linear interpolation or time window-based interpolation strategies are used for ordered scores. When performing standardization, the server obtains the mean and standard deviation of each feature based on the training set statistics, and uses linear transformations to map the features to a zero-mean, unit-variance space, or maps the features to a defined range based on quantiles, thereby reducing numerical instability caused by different units of measurement.

[0111] The server utilizes a feature generation module to construct high-level features based on the original fields. For example, based on performance scores from multiple past evaluation periods, the server calculates the rolling average, rolling standard deviation, and slope (trend strength); based on progress records, the server calculates the on-time completion rate, the distribution of delayed days, and the average task cycle length; based on target weights and the current completion percentage, the server calculates the weighted completion rate and the remaining workload estimate. The server adds these features to the feature data table in column form and records the feature version number to ensure a consistent feature set is used in subsequent model training and inference phases.

[0112] Regarding the prediction model, the server trains a supervised learning model on the processor using a prediction model building program. In one embodiment, the server selects a tree-based ensemble model, such as an ensemble regression model or classification model based on multiple decision trees, to predict whether the goal has been achieved and the performance score. The server extracts the input matrix from the feature data table, using "whether the goal was achieved on time" as the binary classification label and "performance score" as the regression objective, and divides the data into training and validation sets using a sample partitioning strategy. During training, the server uses the objective function as the loss function, and in each iteration, updates the splitting features and leaf node outputs of each tree based on the error between the current model output and the true label, thereby gradually reducing the overall loss. After training, the server serializes the model parameters and stores them in a non-volatile storage device for online inference.

[0113] In another embodiment, the server uses a multilayer perceptron-type neural network as the prediction model. This neural network includes an input layer, multiple fully connected hidden layers, and an output layer; the hidden layers use non-linear activation functions, and the output layer uses either a logistic function or a linear output depending on the task type. During training, the server employs a gradient descent variant to calculate the gradients of each parameter based on an error function (e.g., cross-entropy or mean squared error), and updates the network weights using backpropagation. The server can also employ regularization, batch normalization, and early stopping strategies to prevent overfitting and improve generalization performance. Through the integration of these machine learning modules, the server internally forms a continuous computational chain from feature generation to prediction output, eliminating the need for manual rule setting, thereby improving prediction accuracy and model updability.

[0114] In terms of visualization generation, the server combines the statistical information from the prediction result table and the feature data table to generate a visual data structure. The server can use the query interface provided by the visualization platform to encapsulate relevant data into a data source view and define chart configurations, such as time-series line charts, bar charts, and radar charts. The server stores the generated visualization configuration in the visualization platform and outputs the visualization page to the terminal via the network. After receiving the page, the terminal executes a script in its local browser engine to render the visualization data from the server, thereby graphically displaying time-varying information, target achievement probability information, and group ranking information on the screen.

[0115] In terms of background information generation, the server reads the user's corresponding parsed data, prediction results data, and group statistics from the data storage device based on the user's login identifier or user identifier provided by the terminal. The server constructs a summary through aggregation operations, such as the average performance score for the most recent few periods, the current main goal and its predicted probability of achievement, and the difference relative to the departmental mean. Simultaneously, the server uses a group information processing program to statistically analyze the indicators of other members, generating baseline distribution information (such as quantiles, mean, variance, and distribution curve description). The server calculates the relative evaluation information based on the position of individual indicators within the baseline distribution (such as being in the top percentile or above the mean by several standard deviations). This information is combined into structured background text or tokenized data, providing context for the generation of subsequent prompts.

[0116] Regarding prompt generation, the server maintains a set of instruction rules, which exist in the form of templates and condition sets. For example, when a user focuses on the main goals for the current year, the prompt should include elements such as "current goal description, deadline, current completion status, model-predicted probability of achievement, and comparison with the departmental average." The server fills in template variables based on the user's interaction context and background information to generate prompts in natural language. These prompts serve as input instructions for the generative AI model, precisely specifying the analysis scope and output format. For example, the server can generate the following prompts: "Based on my performance, goals, and progress data recorded in the system, please analyze the probability of my success in achieving the main goals for the current year, and list the three most important influencing factors and suggestions for improvement." For example, the server can generate the following prompt: "Please help me plan a detailed work schedule for the next four weeks based on the task completion status of the past two months, prioritizing the timely completion of high-weight targets." For example, the server can generate the following prompt: "Please compare my performance with the top 10% of employees in this department, point out my weaknesses in skills and work habits, and provide a specific improvement plan." The server concatenates the aforementioned prompts with background information as input to the generative AI model. In this embodiment, the generative AI model can employ a sequence-to-sequence language model based on a self-attention mechanism, containing multiple layers of encoder-decoder stacks, each layer including a multi-head self-attention sublayer and a feedforward sublayer. During model training, the server uses large-scale text corpora and structured data converted to text corpora, training the model parameters through a maximum likelihood objective function, and generating output text during the inference phase using a bundle search or sampling strategy.

[0117] In the generative AI model interface module, the server sends the constructed input text to the inference service via a network request. The inference service performs forward propagation computation on hardware equipped with a graphics processing unit, calculating the output distribution at each time step and generating response text word by word. After receiving the model output, the server performs post-processing on the output text, such as removing redundant prefixes, segmenting by paragraph, detecting and completing incomplete sentences, and storing the generated suggestions and explanatory information in association with the structured prediction results. In this way, the server not only provides natural language suggestions but also references specific numerical predictions and visualization links, allowing users to jump from the text to the corresponding chart views.

[0118] In the end-to-end data flow, this invention achieves a non-traditional processing approach by tightly coupling modules such as data acquisition, feature generation, predictive computation, population distribution statistics, prompt statement construction, and generative AI inference within the server. Unlike reporting systems that rely solely on fixed rules or manual configuration, this invention utilizes a predictive model to automatically learn high-dimensional features. Based on this, the prompt statement generation module constructs inputs that strictly match the background data and feeds them into the generative AI model. This ensures that the latter's output is not merely generalized text, but a precise explanation and executable plan tailored to a specific user, a specific time window, and a specific goal. This linkage from the feature space to the natural language space enables the server to implement a new information encoding and decoding path internally, thereby improving the efficiency of computing resource utilization and the ability to interpret results.

[0119] In terms of technical effectiveness, the server reduces numerical instability and noise interference during model training by standardizing features, filling in missing values, and generating higher-order features, resulting in a significant reduction in prediction error compared to using raw data input. By employing ensemble learning or deep network structures, the server can perform batch inference on a large number of users and targets with less prediction latency under the same hardware, achieving high throughput in prediction processing. The server caches feature data and population statistics at the database level, avoiding repeated high-complexity statistical operations with each user request, thereby reducing disk access and network transmission volume, lowering communication load, and reducing overall response time. Through a unified data model, the server binds prediction results and generative AI outputs to the same data structure, allowing the terminal to obtain numerical predictions and explanatory text with only one round trip upon request, effectively reducing the number of interactions between the client and server.

[0120] In this invention, the terminal serves as an electronic device for user operation, and can be a portable terminal or a desktop terminal. The terminal connects to the server via a network, and displays the visual information and generated results provided by the server on a display device. The terminal executes the script engine of the browser or client application locally, sending query parameters and new prompts to the server based on the user's clicks, inputs, and swipes. The terminal itself does not perform complex prediction and generation calculations; instead, it centralizes the main computational tasks on the server side to utilize the server's high-performance hardware resources and optimized data structures.

[0121] Users primarily play an interactive role in the system of this invention. Users browse visual charts on the terminal to understand current performance and target progress; users read suggestion and explanatory information generated by the generative artificial intelligence model and adjust their work plans or communicate with managers accordingly. Users can modify the prompts according to actual needs, such as adding specific project names or time ranges. The server will then regenerate background information based on the new prompts and trigger a new generative artificial intelligence inference process. Through this flexible interaction method, the technical solution of this invention hides the complex calculation process within the server, enabling users to drive high-dimensional data analysis and personalized suggestion generation in natural language.

[0122] In some alternative implementations, the server can employ different types of prediction models, such as gradient boosting tree models, regression tree models, or convolutional neural network models; the generative artificial intelligence model can also employ different parameter configurations and different training datasets. The server can choose to perform inference operations on the graphics processing unit or only on the central processing unit, depending on hardware resources. At the data storage level, the server can also choose a columnar database or a distributed database to support larger datasets and higher concurrent access volumes. These alternative approaches, without altering the fundamental idea of ​​the invention, can all achieve the integration of multi-source data, feature generation, prediction, and generative artificial intelligence output based on prompts, thereby improving data processing efficiency, prediction accuracy, and the quality of human-computer interaction.

[0123] use Figure 11 The processing flow is explained.

[0124] Step 1: The server obtains structured data from the personnel information processing device, business evaluation processing device, target management processing device, business progress analysis processing device, and learning processing device through communication interfaces.

[0125] Input: Structured data (e.g., JSON or tabular format) provided by each processing unit through a network interface, including basic personnel information, performance ratings, goal settings, progress records, etc.

[0126] The server parses the received structured data based on a preset data field mapping table, identifies key fields such as employee identifier, time period, and indicator name, and reorganizes data from different sources according to a unified internal field naming rule. It also performs data type conversion (e.g., string to timestamp, string to floating-point number) and basic validity checks (e.g., whether required fields are empty and whether values ​​are within a reasonable range).

[0127] Output: A collection of intermediate data records conforming to the internal unified format, stored in the server's memory cache for subsequent integration and processing.

[0128] Step 2: The server integrates intermediate data records from different processing devices based on identifiers and writes them to the data storage device.

[0129] Input: The set of intermediate data records generated in step 1, including multi-source data with fields such as employee identifier, target identifier, and timestamp.

[0130] The server uses employee ID and time period as a composite key to perform association operations on personnel information, performance records, target records, and progress records, constructing a unified employee-time dimension record. For target-related data, the server uses target ID as the key to merge target setting information with progress information. The server performs deduplication on duplicate records and selects conflicting records according to timestamp or priority rules. After processing, the server generates relational table-formatted data and uses a database driver to insert or update the data into the employee information table, performance record table, target record table, and progress record table in the data storage device.

[0131] Output: Multiple basic business data tables persistently stored in the data storage device, serving as the input source for subsequent feature generation and prediction calculations.

[0132] Step 3: The server reads basic business data from the data storage device, executes a data parsing program, fills in missing values, standardizes and summarizes the data, and generates feature data.

[0133] Input: Relational table data such as employee information table, performance record table, target record table, progress record table, etc.

[0134] The server first filters records based on a preset time window (e.g., the last four assessment periods). Then, it calls a data parsing program to check each field for missing values. For continuous fields, it uses the grouped mean or median to complete the data; for categorical fields, it uses the mode or default category to complete the data. Next, the server performs standardization on each continuous indicator, calculating the mean and standard deviation of each indicator and applying a linear transformation to map the data to a uniform scale. Subsequently, the server performs summary calculations on performance, target, and progress data according to employee identification and time period, such as calculating multi-period average performance, cumulative number of completed tasks, and weighted completion rate of the current target.

[0135] Output: A parsed data set containing cleaned indicators and basic statistical summary results, organized into the basic fields of the feature data table.

[0136] Step 4: The server performs feature generation based on the parsed data set, constructs higher-order features, and writes them into the feature data table.

[0137] Input: The parsed data set generated in step 3, including various indicators that have been completed, standardized, and summarized.

[0138] The server performs mathematical and time-series calculations on the parsed data according to predefined feature generation rules. For example, it calculates the rolling average, rolling standard deviation, and linear regression slope based on multi-period performance scores; it calculates on-time completion rate, delay rate, and average task cycle length based on progress records; and it calculates weighted completion rate and remaining workload estimate based on target weights and completion rates. The server appends these calculation results as new feature columns to the data structure, recording the feature name, version information, and calculation time. Subsequently, the server updates the feature data to the feature data table in the data storage device through batch write operations.

[0139] Output: A feature data table containing multi-dimensional high-order features, providing the input feature matrix for the prediction model.

[0140] Step 5: The server constructs a training dataset for the prediction model from the feature data table and trains the prediction model.

[0141] Input: Feature columns and corresponding label columns in the feature data table (e.g., binary labels for whether the goal was achieved, and continuous value labels for performance scores).

[0142] The server selects the feature columns to be used for training based on the configuration, removes invalid or highly correlated redundant features, and forms an input feature vector set; simultaneously, it extracts label values ​​from target records and performance records. The server divides the data into training and validation sets according to a preset ratio, and iteratively trains the training set using machine learning algorithms (such as ensemble tree models or multilayer perceptron models). In each iteration, the server calculates the loss function value based on the error between the current model output and the true label, and updates the model parameters using methods such as gradient or split gain. After each round of training, the server uses the validation set to evaluate the model performance and determines whether to continue training or perform hyperparameter tuning based on metrics (such as accuracy, AUC, or mean squared error). Finally, the server saves the converged model parameters to persistent storage.

[0143] Output: The trained prediction model and its parameter file, which can be used in the online inference stage.

[0144] Step 6: The server uses a trained prediction model to perform online inference on the current data, generate the probability of achieving the target and business efficiency indicators, and writes them into the prediction results table.

[0145] Input: Feature records for targets that have not yet ended and the latest cycle in the feature data table, and the prediction model trained in step 5.

[0146] The server selects samples to be predicted from the feature data table, constructs a prediction input matrix, and feeds the matrix into the prediction model for forward computation. For classification models, the server outputs the probability value of each sample belonging to the "on-time achievement" category; for regression models, the server outputs the predicted performance score or efficiency score. The server combines the prediction output with the corresponding employee identifier, target identifier, and timestamp into a structured record, writes it to the prediction results table, and can record the model version number for tracking.

[0147] Output: A table of prediction results containing the probability of goal achievement, predicted performance scores, and efficiency indicators, providing input for subsequent visualization and generative AI analysis.

[0148] Step 7: The server generates visualized data and configures visualized views based on the parsed dataset and predicted results.

[0149] Input: Parse the dataset, the prediction results table, and group statistics (such as departmental averages and distribution information).

[0150] The server calculates each employee's time-series metrics, the probability of achieving current key objectives, and departmental ranking through aggregation operations. It then organizes these values ​​into chart-like structures, such as time-series lists, ranking lists, and radar chart coordinates. The server registers this data as a data source for the visualization platform and creates corresponding chart configurations (such as line charts, bar charts, and heatmaps) for each analysis scenario. The server stores the generated visualization configurations and data source bindings within the visualization platform, enabling a graphical interface to be provided to terminals via web pages or client applications.

[0151] Output: A visual description of the data source and chart configuration, which is loaded by the terminal and displayed on the display device.

[0152] Step 8: The terminal requests a visualization page and related data from the server and renders and displays it locally.

[0153] Input: The address of the visualization page and the visualization data source interface provided by the server.

[0154] The terminal accesses the server's page interface via the network, receiving HTML pages, style files, and script files. Subsequently, the terminal executes the scripts in its local browser engine, calling the server's data interface to obtain specific visualization data. The terminal maps time-series data, probability data, and ranking data to graphical elements (such as coordinates, bar heights, and color depths), drawing line charts, bar charts, radar charts, etc., on the display screen. Users filter, zoom, and switch views on the terminal using touch or mouse operations. Based on user actions, the terminal sends new query parameters to the server to obtain updated visualization data.

[0155] Output: A dynamic graphical interface displayed on the terminal screen, providing an intuitive background for subsequent user interactions and input prompts.

[0156] Step 9: After viewing the visualized information on the terminal, users can input natural language prompts according to their needs to request generative artificial intelligence analysis.

[0157] Input: The visual interface displayed on the terminal and the personal performance, target progress, and relative ranking information shown therein.

[0158] Users input prompts into a dialog box using the terminal's input device (keyboard, touchscreen), such as: "Based on my performance, goals, and progress data recorded in the system, please analyze the probability of my success in achieving my main annual goals, and list the three most important influencing factors and improvement suggestions." or "Please help me plan a detailed work schedule for the next four weeks based on my task completion status over the past two months, prioritizing the timely completion of high-weight goals." The terminal packages the user's input text along with the user identifier into a request message and sends it to the server's generative artificial intelligence interface.

[0159] Output: A request message containing the user identifier and prompt statement, submitted to the server to initiate subsequent background information generation and generative AI inference processing.

[0160] Step 10: The server retrieves relevant data from the data storage device based on the user identifier and generates background information.

[0161] Input: The user identifier and prompt statement sent by the terminal in step 9.

[0162] The server retrieves the user's performance records for the most recent periods, current main goals and progress, predicted achievement probability, and relative ranking within the department from the parsed data set and prediction results table based on the user's identifier. Simultaneously, it obtains the baseline distribution information for the corresponding department from the group statistics table. The server then aggregates and formats this data into textual background information, such as: "The most recent 4 performance ratings are...; the current annual main goal is..., the current completion rate is..., the predicted achievement probability is...; the department's average rating is..., and you are currently in the top...%."

[0163] Output: Background information text associated with the user and the current analysis context, providing context for prompt generation and generative AI input construction.

[0164] Step 11: The server generates prompts for generative artificial intelligence models based on background information and instruction rules, and constructs input information.

[0165] Input: The background information text generated in step 10, as well as the original user prompts and internal system instructions.

[0166] The server first parses the user prompt to identify the type of analysis requested (e.g., success probability analysis, task planning, gap analysis). Then, based on the corresponding instruction rules, it determines the necessary background fields and output format requirements. The server combines the background information fragments with the user prompt to form a complete prompt, adding system instructions and output requirements beforehand, such as specifying "Please answer in Simplified Chinese" or "Please output in bullet points." Subsequently, the server uses this complete text as input to the generative AI model, preparing to send it to the inference service.

[0167] Output: A complete textual prompt containing system descriptions, background information, user requirements, and output format requirements; that is, the input information for the generative artificial intelligence model.

[0168] Step 12: The server calls the generative artificial intelligence model inference service to generate suggested and explanatory information based on the input information, and then integrates the output.

[0169] Input: The complete prompt text constructed in step 11.

[0170] The server sends the input information to the generative AI model inference service via a network request. The inference service performs forward propagation computation within the model, generating response text sequentially. After receiving the model output, the server segments and formats the text, such as dividing it into segments, numbering it, and marking it with a description of the success probability, a list of key influencing factors, and specific suggestions. The server associates the generated suggestions and explanatory information with the numerical results in the prediction results table, and writes the integrated generated results, along with metadata such as timestamps and model version numbers, into the generated results table.

[0171] Output: The generated result records are stored in a structured table, along with natural language suggestion text that can be requested by the terminal.

[0172] Step 13: The terminal retrieves the generated results from the server and displays the analysis conclusions and recommendations to the user on the screen.

[0173] Input: The query interface for generated results provided by the server and the identifier of the generated result record.

[0174] The terminal requests the latest generated results from the server, receiving text content containing suggestions and explanatory information, as well as associated numerical data such as predicted probabilities. The terminal displays this text content in a dialog-like or report-like format on its local interface, and can embed links within the text, allowing users to click and jump to the corresponding visualization chart page.

[0175] Output: A generative AI analysis results interface displayed on the terminal screen, allowing users to understand their own situation and adjust their behavior accordingly or ask further questions.

[0176] 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".

[0177] In traditional human resource management and business progress management technologies, while servers can collect employee, target, and business progress information from multiple business subsystems, they typically only present static results to users in the form of reports or simple visualizations, lacking the ability to intelligently and personally intervene based on real-time data. Specifically, existing technologies have the following problems: (1) The servers simply aggregate multi-source data from human resource management systems, performance evaluation systems, target management systems, and business progress management systems. They lack a mechanism for deep feature extraction and structured modeling of these heterogeneous data, making it difficult to form high-quality inputs that are instructive for individual work scenarios. As a result, they cannot fully leverage the capabilities of generative artificial intelligence models in understanding complex tasks and generating strategies.

[0178] (2) Even if the server introduces machine learning or rule engine, it is mostly limited to offline scoring, KPI statistics and other purposes. It fails to construct prompts that are strongly related to specific business contexts for generative artificial intelligence models, which makes the model generation results disconnected from the actual working environment. This results in insufficient executability and real-time performance of the generated content, and cannot effectively guide front-line work behavior.

[0179] (3) Existing systems generally output intelligent analysis results through terminal application interfaces or large screens, with coarse output granularity and no optimization for the display limitations of wearable visual display terminals. For example, long text suggestions are not broken down into short text steps adapted to small display areas and limited attention spans, and no display order and status control information is attached, making it difficult for users to obtain key information in a timely manner when working on the go or when their hands are occupied.

[0180] (4) Existing human-computer interaction paths are usually one-way “server output - user passive viewing”, lacking a mechanism to collect user’s execution behavior, interaction operations and timestamps of each step in real time through wearable terminals and to send the log data back to the server in a closed loop. As a result, it is impossible to incorporate the user’s real execution path and feedback into the data set at the system level, making it difficult to form a continuous self-optimization capability for specific scenarios.

[0181] (5) For relative evaluation and incentives at the team level, existing technologies often only rank or score based on offline aggregated indicators. They lack the ability to use generative artificial intelligence models to perform semantic hierarchical analysis and comparison of multi-member progress data, making it difficult to generate relative evaluation information that can be dynamically presented in front-line operations in a timely manner. This affects the refined incentive and guidance effect achieved based on computer technology.

[0182] Therefore, it is necessary to provide a new system and its computer implementation method, which can perform structured processing and feature extraction of multi-source human resource data and business progress data on the server side, automatically construct high-quality prompt statements, drive generative artificial intelligence models to generate work steps and suggestions that are highly matched with specific work scenarios, and present them in a fine-grained and interactive manner through wearable visual display terminals. At the same time, the user's execution logs are fed back to the server, forming a closed loop of data-model-interaction, thereby improving the real-time performance, intelligence and executability of human resource management and business progress management systems at the computer technology level.

[0183] 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.

[0184] In this invention, the server includes: a processing component for acquiring and integrating human resource information, business progress information, and target information from multiple information sources; a data analysis component for calculating statistics and features based on the integrated data set; a prompt construction component for converting the statistics and features into prompt statements suitable as input to a generative artificial intelligence model; a model inference component for calling the generative artificial intelligence model and generating work steps or suggestions related to improving work efficiency based on the prompt statements and the data set; a result arrangement component for segmenting the generated results into small-granularity short texts adapted to wearable visual display terminals and attaching display order information to form terminal-side display data; and a log closed-loop component for receiving user interaction logs from wearable visual display terminals and appending the logs as new human resource information and business progress information to the integrated data set. This allows for the formation of a complete technical chain within the computer, from the automatic structuring of multi-source data, the automatic generation of prompts, the reasoning of generative artificial intelligence models, the fine-grained formatted display of results for wearable terminals, to the automatic feedback of user behavior logs and their participation in subsequent feature updates and prompt optimization. This enables a holistic improvement in the data processing and human-computer interaction processes, thereby enhancing the system's ability to analyze real-time business progress and dynamically guide individual work processes. Ultimately, this improves the performance and effectiveness of human resource management and business progress management at the computer technology level.

[0185] "Human resources information" refers to structured or semi-structured data related to personnel identity, job attributes, competency evaluation, performance records, etc., used to characterize an individual's role and historical performance in an organization.

[0186] "Business progress information" refers to data related to task status, process execution, start time, expected end time, current completion percentage, error records, etc., used to characterize the execution process and real-time progress of business activities.

[0187] "Target information" refers to data related to task objectives, performance targets, deadlines, priorities, weights, etc., used to characterize the expected results that a system or personnel should achieve within a certain period.

[0188] A “dataset” refers to one or more datasets that are formed by integrating, cleaning, and standardizing the format of human resources information, business progress information, and target information obtained from multiple information sources, and are used for subsequent analysis and modeling.

[0189] "Statistics" refer to descriptive indicators calculated based on a dataset, including but not limited to mean, variance, standard deviation, maximum, minimum, frequency, proportion, etc., used to characterize the distribution characteristics of data.

[0190] "Features" refer to the numerical or symbolic representations extracted or constructed from a dataset and used as input to generative artificial intelligence models or other algorithmic models, including original fields, derived fields, and encoded features.

[0191] "Generative artificial intelligence models" refer to models trained through machine learning that can automatically generate text, instructions, or other content based on input data and prompts. They are typically implemented using deep learning structures and are used to output task steps, suggestions, or evaluation results.

[0192] "Prompt statements" refer to natural language or structured text constructed by the server based on statistics, features, and business context, which are used as input to generative artificial intelligence models to instruct the model to perform specific analyses or generate specific types of content.

[0193] "Generated results" refers to the text content or other forms of results output by the generative artificial intelligence model after receiving prompts and related data. In this invention, it mainly includes work steps, suggestions and evaluation indicators related to improving work efficiency.

[0194] "Terminal-side display data" refers to structured data generated by the server based on the generated results to adapt to the display characteristics of wearable visual display terminals. It includes at least segmented short text content, display order information, and additional information related to display control.

[0195] "Wearable visual display terminal" refers to a terminal device that can be worn by a user and present visual information in their field of vision, including head-mounted display devices, smart glasses, etc., used to display work steps, suggestions and evaluation information in real time during the work process.

[0196] "User input operations" refer to interactive behaviors performed by users through wearable visual display terminals, including touch, button, voice commands, gestures, etc., used to switch displayed content, confirm the completion of steps, or trigger other interactive events.

[0197] "Log information" refers to the data recorded on wearable visual display terminals based on the user's work steps and input operations, including step number, operation type, timestamp, task identifier, etc., which is used to reflect the user's actual execution and interaction process.

[0198] "Relative evaluation index" refers to the quantitative evaluation result that is calculated by generative artificial intelligence models or other algorithms based on the human resources information and business progress information of multiple members, and is used to represent the relative position or performance level of the target user in the group.

[0199] The "result orchestration component" refers to a functional module that runs on the server side, splits, reorganizes, formats, and adds sequence control information to the generated results output by the generative artificial intelligence model to form data displayed on the terminal side.

[0200] The "log closed-loop component" refers to a functional module that runs on the server side, used to receive log information from wearable visual display terminals and integrate the log information into a dataset for subsequent statistical updates, feature extraction, and optimization of prompt statements.

[0201] In one embodiment, the server includes one or more computing devices, such as a central processing unit (e.g., a multi-core general-purpose processor), a graphics processing unit (e.g., a parallel processing chip for deep learning inference), semiconductor memory, and non-volatile storage media. The server runs background service programs, database management programs, and generative artificial intelligence inference services on an operating system (e.g., a UNIX-based server operating system). In one embodiment, the terminal includes a wearable visual display terminal, such as a head-mounted display device or smart glasses with a microdisplay, embedded processor, wireless communication module, and input sensors, and runs client applications on an embedded operating system (e.g., a mobile terminal operating system or a custom real-time operating system). In one embodiment, the user performs assembly, inspection, and handling tasks in a real-world work environment while interacting with the terminal.

[0202] In one embodiment, the server stores the program in a non-volatile storage medium. When the program is executed on the central processing unit, the server functions as multiple functional modules, including a human resources information acquisition module, a business progress information acquisition module, a target information acquisition module, a data integration module, a feature extraction module, a prompt statement construction module, a generative artificial intelligence model inference module, a result arrangement module, and a log closure module. In another embodiment, the terminal stores the client program in a local storage medium. When the program is executed on the terminal processor, the terminal functions as a communication module, a display control module, a user input acquisition module, and a local buffer module.

[0203] In one implementation, the server uses a database management system (e.g., a relational database management system) to manage human resource information, business progress information, and target information. In the data integration module, the server reads records from multiple logical tables using structured queries and employs an intermediate data structure (e.g., record objects organized by key-value pairs or a two-dimensional table structure) to uniformly map and convert field names such as employee identifier, task identifier, task type, start time, expected end time, and historical performance indicators. In the feature extraction module, the server uses data analysis libraries (e.g., numerical computation libraries and data frame processing libraries) to calculate statistics such as average execution time, standard deviation of execution time, error rate, and execution frequency from centrally stored historical records. It further constructs combined features, such as the difference between historical average execution time and current expected execution time, performance ratios across different time periods, and frequency vectors of common error types. The server uses these statistics and features to form numerical vectors suitable for machine learning input and stores them in a specific order in a feature table for fast indexing and batch reading.

[0204] In one implementation, the server transforms the prompt statement construction module into a template-based text generation engine. Within this module, the server embeds the aforementioned statistics, features, and current task status into a predefined natural language template, automatically generating grammatically correct prompt statements that contain key information. For example, the server generates the following prompt statement: "You are an expert in optimizing production line efficiency."

[0205] An operator is currently assembling electronic components.

[0206] The operator's historical data is as follows: - Task type: Component assembly - Average time spent on this type of task over the past 3 months: 20 minutes - Standard deviation of time: 4 minutes - Average error rate: 5% The current task data is as follows: - Current progress: 60% - Estimated total time: 25 minutes Objective: To improve the efficiency of remaining processes without increasing the error rate.

[0207] Please generate three concise efficiency improvement suggestions that can be directly displayed on your smart glasses. Each suggestion should not exceed 20 Chinese characters, and output them in the format of 'Suggestion 1:…'. In another implementation, the server generates the following prompt statement: "Based on the following task progress and historical performance, please design an optimized sequence of subsequent tasks."

[0208] Require: 1. Only output subsequent steps, without describing what has been completed; 2. Each step is described in one sentence, making it suitable for displaying step-by-step on smart glasses; 3. The steps are numbered using 'Step 1: ..., Step 2: ...'.

[0209] data: Current progress: 60%; Remaining steps: Install component A, test component B, and finally tighten. Historical average time: 20 minutes, current estimated time: 25 minutes; Error rate: 5%.

[0210] Please provide the optimized step sequence and precautions for each step. The server runs a generative AI model within the generative AI model inference module. In one implementation, the server employs a sequence-to-sequence neural network structure based on a self-attention mechanism, comprising an embedding layer, a multi-layer encoder, a multi-layer decoder, and an output layer. During training, the server uses a large number of anonymized historical task records and corresponding suggested texts as training samples to optimize model parameters. During training, the server uses cross-entropy loss as the objective function and updates parameters using gradient descent and its variants (such as adaptive learning rate optimization algorithms). In each training round, it calculates the difference between the output distribution and the target text distribution and backpropagates the error to update network weights. During training, the server performs data augmentation operations on the input data, such as synonym replacement, field order perturbation, and noise injection, to enhance the model's robustness to diverse text expressions in real-world business environments.

[0211] During the inference phase, the server receives prompts and first converts the text into a token sequence through word segmentation and encoding, then transforms it into a vector representation through an embedding layer. In the encoder, the server performs multi-head self-attention computation and feedforward network operations to generate a hidden representation containing contextual information. Subsequently, in the decoder, subsequent tokens are generated progressively based on previously generated tokens and the encoder output. The server performs probability distribution calculations on the vocabulary at each decoding step, selecting the token with the highest probability or generating the next token according to a predefined sampling algorithm, until a termination token is generated or a predetermined length is reached. In this way, the server generates task steps or suggested text tailored to specific tasks and employee characteristics.

[0212] In the results orchestration module, the server performs post-processing on the long text returned by the generative AI model. The server executes a text segmentation algorithm, splitting the entire text into multiple shorter texts based on patterns such as "Suggestion 1:", "Suggestion 2:", or "Step 1:", "Step 2:". The server calculates the character length of each short text, truncating it or generating a shorter expression by re-invoking the model if it exceeds the single-screen display limit of the wearable visual display terminal. The server assigns a sequence number, priority, and display duration parameter to each short text, organizing it into a terminal-side display data structure according to a fixed format. The server saves the display parameter configuration in non-volatile storage media, allowing adjustment of the maximum number of characters per screen and refresh rate for different terminal models, improving display efficiency and reducing the rendering burden on the terminal side, thereby reducing terminal processing power consumption.

[0213] The server receives log information from the terminal in the log closed-loop module. The server parses the logs, including fields such as step number, operation type, and timestamp, and writes them to the log table. Based on the log data, the server calculates the actual execution time distribution and the user's acceptance of the suggested steps, and periodically updates statistics and features. In this way, the server ensures that subsequent prompts and model inputs reflect the latest job behavior patterns, gradually correcting prediction bias and suggestion failures, achieving data-driven adaptive optimization. Because the server directly uses log data when updating features and constructing prompts, it can continuously improve the accuracy of generated content without manual labeling, thereby reducing prediction errors and shortening model convergence time at the computational level.

[0214] The terminal establishes a persistent connection with the server via a wireless communication interface in its communication module. The terminal receives display data from the server using a lightweight messaging protocol and stores the data structure corresponding to the current task in its local memory. In its display control module, the terminal displays work steps or suggestions sequentially on a micro-display based on the received display order information and priority. In one embodiment, the terminal uses a hardware-accelerated graphics library to render text, reducing rendering latency; in another embodiment, the terminal automatically adjusts the display brightness and contrast based on ambient light sensor data to ensure clear text display under varying lighting conditions.

[0215] The terminal receives user input commands via touch, button, or voice input through its user input acquisition module. It associates each operation with the currently displayed step number, generating a log entry with a timestamp. The terminal temporarily stores these log entries in a local cache module to handle network fluctuations, and uploads them to the server in batches when the network connection is stable. This batch upload strategy reduces the number of communication requests, thereby lowering communication load and energy consumption, and improving the overall system response time.

[0216] In a specific usage scenario, a user wears a terminal to perform component assembly tasks on a production line. At the edge of their field of vision, the user sees short text prompts on the terminal, such as "Current Progress: 60%" and "Step 1: Install Component A first to reduce interference." Following these prompts, the user adjusts the original work sequence, first completing the installation of larger components, then simultaneously preparing tools for subsequent testing during the installation process, and finally completing the tightening step after all components are in place. After completing each step, the user taps the terminal's touch area to indicate "Complete current step," and the terminal then switches to displaying the next suggestion. This collaborative approach reduces the number of tool changes and unnecessary actions, shortens the overall task time, and lowers the error rate.

[0217] In one variant implementation, the server not only generates work steps and suggestions, but also constructs prompts for calculating relative evaluation metrics based on the progress information of other team members. For example, the server generates the following prompt: "Based on the average time and error rate of the following multiple operators on the same task, compare the difference between the current operator and the average level of the team, and output a concise relative evaluation result."

[0218] Require: 1. Indicate whether the current operator is better than, worse than, or close to the average level in terms of time consumption and error rate; 2. The output content should be displayed as one or two sentences on smart glasses.

[0219] data: Current operators: average time spent 22 minutes, error rate 4%; Team average: average time spent 20 minutes, error rate 5%.

[0220] Please provide the relative evaluation results directly. The server extracts relative evaluation metrics from the text output of a generative artificial intelligence model and adds them as additional fields to the data displayed on the terminal. This allows the terminal to display the user's real-time position relative to the team benchmark while showing task suggestions. By encapsulating the complex multi-member data comparison process into the model inference and prompt statement construction process, the server eliminates the need for complex multi-dimensional calculations on the terminal, thereby reducing the terminal's computational burden and improving the overall system processing efficiency.

[0221] The server achieves scalability across multiple implementations through a modular structure. For example, the generative AI model inference module can be replaced with language models of varying sizes, the prompt construction module can employ rule-based templates or learnable prompt generation networks, and the results arrangement module can adjust its splitting and sorting strategies for different terminal display specifications. Furthermore, the server can be configured with different feature selection algorithms in different implementations, such as using gradient boosting tree-based importance ranking or sparse regularization-based feature selection methods to reduce input dimensionality, decrease model inference time, and lower memory consumption.

[0222] Through the aforementioned technical solutions, the server internally organizes multi-source information using a specific data structure and achieves higher expressive power and adaptability compared to traditional rule-based systems through a combination of feature extraction, prompt generation, and generative artificial intelligence model inference. The terminal, through short text arrangement and batch log uploading strategies tailored to wearable visual display terminals, achieves efficient display and low-load communication on resource-constrained devices. Users, through interaction with the terminal, not only passively receive information but also actively provide the server with computationally usable execution data, forming a self-reinforcing closed loop of data-model-interaction. These specific hardware configurations, software module divisions, data structure designs, and algorithmic flows enable this system to go beyond abstract business management logic, achieving technical improvements at the levels of internal computer storage structure, computational processes, and communication protocols. This results in enhanced processing speed, improved prediction accuracy, reduced communication load, and improved data management efficiency.

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

[0224] Step 1: The server retrieves and integrates data from multiple information sources.

[0225] Inputs: Human resources information from the Human Resources Management database, business progress information from the Business Progress Management database, and target information from the Target Management database.

[0226] The server uses database drivers and structured query statements to read fields such as employee ID, task ID, task type, evaluation record, start time, estimated end time, current progress percentage, and target description from different logical tables. The server performs a mapping of field names across different systems, unifying fields with the same meaning into standard names, converting date strings to timestamps, and converting text-based numeric values ​​to numeric values. The server then combines the cleaned records into a unified intermediate data structure (e.g., record objects with employee ID and task ID as keys) and writes them into a unified data set table.

[0227] Output: A unified data set with consistent field formats and standardized data types.

[0228] Step 2: The server calculates statistics and features from the integrated dataset.

[0229] Input: The integrated data set output from step 1.

[0230] The server uses a data analysis library to group and calculate the historical records of each employee across various tasks, obtaining statistics such as average time, standard deviation of time, average error rate, and number of task completions for each task type. The server further constructs feature quantities based on these statistics, such as the difference between the current estimated time and the historical average time, the ratio of the current number of errors to the historical average number of errors, and the frequency vector of specific error types. The server encodes these feature quantities into numerical feature vectors and stores them in a feature table, associating them with the corresponding employee and task identifiers.

[0231] Output: A set of feature data containing statistics and features.

[0232] Step 3: The server provides prompts for constructing generative artificial intelligence models.

[0233] Input: The feature data set output from step 2 and the current task status information.

[0234] The server calls string templates in the prompt statement construction module, embedding statistics and features into predefined templates in text form to generate natural language prompt statements. For example, the server inserts fields such as "average time 20 minutes," "current estimated 25 minutes," "error rate 5%," and "current progress 60%" into multi-line Chinese descriptions, adding task type and goal descriptions to form complete prompt statements. When multiple output formats are needed, the server selects different templates to generate multiple prompt statements; for example, one for generating efficiency improvement suggestions and another for generating work step sequence optimization schemes.

[0235] Output: Prompt text for generative artificial intelligence models.

[0236] Step 4: The server calls a generative artificial intelligence model to generate text results.

[0237] Input: The prompt statement output in step 3 and the context parameters related to the task (such as maximum generation length, temperature parameters, etc.).

[0238] The server feeds the prompts into the generative AI model inference service. First, it performs word segmentation and encoding on the prompts, converting the text into a token sequence, which is then mapped into vectors through an embedding layer. In the encoder, the server performs multi-head self-attention and feedforward network operations to generate a contextual representation; in the decoder, it progressively outputs the next token based on the prompts and previously generated tokens. At each decoding step, the server calculates the probability distribution over the vocabulary, selecting the token with the highest probability or sampling according to a set strategy until a termination token is generated or the length limit is reached. Finally, the server decodes the token sequence into Chinese text, such as several phrases like "Suggestion 1: ..., Suggestion 2: ..." or "Step 1: ..., Step 2: ...".

[0239] Output: Raw generated text results related to job efficiency improvements.

[0240] Step 5: The server splits and formats the generated results.

[0241] Input: The original generated text result from step 4.

[0242] The server uses a text processing module to identify pattern markers (such as "Suggestion 1:", "Step 1:", etc.) in the generated text, and segments the text into multiple short sentences according to the marker positions. The server counts the number of characters in each short sentence, truncating sentences that exceed the single-screen display limit of the terminal or generating simplified descriptions by calling the generative artificial intelligence model again. The server assigns a sequence number, priority, and suggested display duration to each short sentence, and merges these fields with employee identifiers and task identifiers to construct the terminal-side display data structure for subsequent transmission.

[0243] Output: Terminal-side display data containing multiple short texts and display control information.

[0244] Step 6: The server sends the data displayed on the terminal side to the terminal.

[0245] Input: The terminal-side display data and terminal identification information output in step 5.

[0246] The server establishes a session connection with the terminal through the communication module, and packages the data displayed on the terminal side using a message protocol, including the task ID, current progress text, short text list and its order information, etc. Before sending, the server performs an authentication check to confirm the terminal's legitimacy, and then sends the data to the corresponding terminal through the network interface. The server records the sending time and message identifier for fault recovery or retransmission control.

[0247] Output: Display data packets successfully transmitted to the terminal, and sending log records on the server side.

[0248] Step 7: The terminal receives and parses the data displayed on the terminal side.

[0249] Input: The display data message sent by the server in step 6.

[0250] The terminal listens for server messages in its communication module. When a message is received, it first verifies the message integrity, then calls a JSON parsing or equivalent structure parsing function to parse the message into an internal data object. The terminal reads fields such as the current progress text, short text array, and display sequence number, stores them in its local memory structure, and initializes the current display index to the first short text. The terminal builds a fast index for frequently accessed fields to improve the speed of subsequent interface refreshes.

[0251] Output: The data structure to be displayed and the initial display state stored in the terminal memory.

[0252] Step 8: The terminal displays work steps or suggestions on a wearable display screen.

[0253] Input: The data structure to be displayed and the current display index output in step 7.

[0254] The terminal calls the graphical interface library in the display control module to draw the current progress text above the display area and the short text corresponding to the current index in the main display area. The terminal sets the font size, line spacing, and refresh interval according to the display control information to ensure the text is clearly readable within the limited space. After displaying the text, the terminal retains the current text for a preset time while listening for user input events to switch to the next or previous short text.

[0255] Output: A screen displaying current task suggestions or steps presented in the user's view.

[0256] Step 9: Users perform actual tasks based on the prompts displayed on the terminal.

[0257] Input: The current job step or suggestion displayed on the terminal in step 8.

[0258] Users read the text displayed on the terminal, such as "Step 1: First install component A to reduce interference," and then perform the corresponding physical actions in the real environment: picking up the specified component and installing or adjusting it as required. After completing the current step, users can send an operation signal by lightly touching the terminal's touch area or pressing a button, indicating that the current step is complete and they are ready to view the next step.

[0259] Output: The actual operation status after completing the corresponding physical operation steps, and the user's input operation signals to the terminal.

[0260] Step 10: The terminal collects user input and updates the displayed content.

[0261] Input: The operation signals generated by the user through touch, key or voice in step 9, and the currently displayed index.

[0262] The terminal identifies the operation type (such as "Next", "Previous", "Confirm Complete") in the user input collection module and records the current timestamp and current step number. The terminal updates the current step's status from "Displaying" to "Completed", and then adjusts the current display index based on the operation type, such as incrementing or decrementing it. The terminal calls the display control module to redraw the interface and display the new short text. Simultaneously, the terminal adds a new log record in its local buffer, containing the step number, operation type, and timestamp, for subsequent uploading.

[0263] Output: The updated current display index, the updated interface display, and newly added local log entries.

[0264] Step 11: The terminal uploads log information to the server in batches.

[0265] Input: The set of local log entries output in step 10 and the terminal and task identification information.

[0266] The terminal packages accumulated log entries into a single log data set in its local buffer module according to a preset strategy (e.g., reaching a certain number of entries or after a certain time). The terminal establishes or reuses a connection with the server via its communication module and sends the log data to the server's designated interface. Upon receiving confirmation from the server, the terminal marks successfully uploaded log entries as "synchronized" to free up local storage space; if the upload fails, the terminal retains the logs for retrying.

[0267] Output: Log data packets uploaded to the server and the updated log buffer status on the terminal side.

[0268] Step 12: The server receives logs and updates the dataset and features.

[0269] Input: The log data message sent by the terminal in step 11.

[0270] In the log closed-loop module, the server parses messages, extracting information such as employee identifier, task identifier, step number, operation type, and timestamp, and stores it in the log table. Based on the logs, the server calculates the actual time spent on each step (such as the difference between adjacent "complete" times) and the user's response behavior to suggestions (such as skip rate and dwell time), and updates historical statistics, such as average time spent, distribution variance, and error association patterns, according to this new data. The server writes the latest statistics back to the feature table and references these updated features when constructing subsequent prompts, thereby adjusting the input distribution of the generative artificial intelligence model.

[0271] Output: An updated dataset containing the latest execution data and a feature dataset.

[0272] Step 13: The server generates relative evaluation prompts based on team data and updates the displayed content.

[0273] Input: The updated data set output from step 12, the feature data set, and the historical and current progress data of other members.

[0274] The server selects statistics from the database for multiple members related to the current task type, calculating the team's average time and average error rate. In the prompt statement construction module, the server embeds the difference between the current user's and the team's average times into a new prompt statement template, such as describing information like "The current operator's average time is 22 minutes, and the team's average time is 20 minutes," and requests the generative AI model to output brief relative evaluation text. The server then re-organizes the relative evaluation results output by the generative AI model along with the original task step suggestions into data for terminal display, and resends and displays it according to steps 6 to 8, allowing users to see both individual suggestions and team comparison information in subsequent tasks.

[0275] Output: Updated terminal-side display data containing relative evaluation information and a new round of display content for the terminal.

[0276] 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.

[0277] 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."

[0278] In the fields of human resource management and performance evaluation, as the number of evaluated entities and evaluation dimensions increase, traditional computer-based implementation methods often suffer from the following technical problems: First, servers typically only perform simple statistics or rule-based scoring on raw evaluation data, lacking a unified and stable standardized processing flow. This makes it difficult to compare evaluation results from different times and groups on the same scale, affecting the reliability of subsequent automated processing. Second, when generating prompts for generative AI models, servers often simply concatenate natural language templates with a small amount of summarized results. The prompts lack structured, modeled relative evaluation indicators, preventing generative AI models from fully utilizing the underlying data and resulting in unstable analysis results and poor interpretability. Third... The server lacks a dedicated calculation mechanism for "relative pseudo-evaluation," and can generally only provide absolute scores or single-dimensional rankings for individuals. It cannot efficiently extract statistical features such as relative position and quantile from large-scale group data in a unified data processing pipeline, and cannot closely integrate them with the reasoning process of generative artificial intelligence models. Fourth, existing systems often implement data analysis modules and generative artificial intelligence model calling modules separately, lacking automated linkage mechanisms for time series and group structures. This makes it impossible to form an integrated processing flow on the server side of "data preprocessing → relative pseudo-evaluation calculation → prompt statement construction → generative artificial intelligence model calling → visualization output," resulting in low system resource utilization efficiency and difficulty in providing users with high-frequency, dynamically updated analysis results in a timely manner.

[0279] In summary, there is an urgent need in the existing technology for an improved computer implementation method that enables the server to embed statistical indicators into prompt statements in a structured manner based on the standardized preprocessing of evaluation data and the modeling of relative pseudo-evaluations. This would stably drive the generative artificial intelligence model and uniformly map the model output and the relative pseudo-evaluation results into visualized data, thereby improving the accuracy and interpretability of data processing while enhancing the automation and computational efficiency of the overall system.

[0280] 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.

[0281] In this invention, the server includes: a processing unit for acquiring evaluation information related to the evaluated object from an information storage device; a processing unit for performing preprocessing operations on the evaluation information, including missing value imputation, outlier removal, and statistical standardization, to generate standardized evaluation data on a uniform scale; a processing unit for performing neighborhood search-based classification operations based on the standardized evaluation data and evaluation data of other objects within the group, calculating relative pseudo-evaluations for each object, and simultaneously calculating overall statistical indicators of the group from the standardized evaluation data and deriving difference indicators, ranking indicators, and quantile indicators; a processing unit for embedding the relative pseudo-evaluations and the statistical indicators in a structured form into prompt statements of a generative artificial intelligence model and generating prompt statements containing original evaluation data and relative pseudo-evaluation information; a processing unit for calling the generative artificial intelligence model based on the prompt statements to perform analysis or prediction processing, and associating the model output results with the relative pseudo-evaluations and the statistical indicators to generate a unified data structure for visualization; and a processing unit for sending the visualization data to a user terminal, simultaneously presenting the relative pseudo-evaluations and the output results of the generative artificial intelligence model in graphical and textual form. This allows for the formation of an end-to-end processing pipeline for evaluation data within the server. At the lower level, standardization and relative pseudo-evaluation modeling improve the stability and comparison accuracy of data processing. In the middle, structured prompts are designed to fully utilize group statistical characteristics to enhance the inference quality of generative artificial intelligence models. At the upper level, a unified visual data structure improves the consistency and interpretability of results presentation, thereby substantially improving the computer's data processing and intelligent analysis capabilities in large-scale evaluation scenarios.

[0282] "Information processing device" refers to a collection of electronic computing devices that include at least one processor and a memory, and are used to execute program instructions to acquire, process, store and control the output of input data.

[0283] "Processing unit" refers to the program logic or functional module executed by the processor in an information processing device, which is used to realize the reading, operation, judgment and control of other components.

[0284] "Information storage device" refers to storage resources used to store evaluation information, model parameters and intermediate processing results in a readable and writable manner, including semiconductor memory, magnetic storage medium or optical storage medium, etc.

[0285] "Evaluation information" refers to raw or processed data related to the behavior, business results, or capability characteristics of the evaluated object, used to reflect the object's performance on specific indicators.

[0286] "Preprocessing" refers to a series of processing operations performed on raw data before modeling or analysis to improve data quality and enhance computational stability, including but not limited to missing value imputation, outlier removal, and standardization.

[0287] "Missing data imputation" refers to the process of estimating and assigning values ​​to missing data items in evaluation information using statistics or rules, so that the data records form a complete vector on each indicator.

[0288] "Outlier removal" refers to the process of deleting, truncating, or replacing values ​​that are determined to deviate significantly from the overall distribution based on predetermined statistical rules or thresholds, in order to reduce the impact of extreme data on subsequent calculations.

[0289] "Standardization based on statistics" refers to using statistical measures such as mean and standard deviation to perform linear transformation on various evaluation indicators so that they meet uniform scale or distribution characteristics, thereby facilitating the comparison of data from different indicators and different periods.

[0290] "Standardized evaluation data" refers to a set of evaluation data that has been preprocessed and unified on a numerical scale, and can be directly used for subsequent modeling and statistical analysis.

[0291] "Group constituent elements" refers to the individual objects being evaluated within the same group or set, including but not limited to personnel, departmental units, or other entities under unified management.

[0292] "Neighborhood search-based classification processing" refers to the calculation process of classifying or classifying target samples based on their distance or similarity in the feature space and by utilizing the label information of their neighboring samples.

[0293] "Relative pseudo-evaluation" refers to an evaluation result calculated based on the evaluation data of other elements within the same group, which is used to represent the relative position or relative level of a certain element.

[0294] "Group overall statistical indicators" refer to statistical quantities used to characterize the overall distribution characteristics, calculated based on standardized evaluation data of a set of group constituent elements. These include mean, variance, percentiles, maximum and minimum values.

[0295] The "difference index" is a quantitative indicator obtained by calculating the difference between an individual's evaluation value and the overall statistical index of the group. It is used to represent the degree of deviation of an individual from the overall level.

[0296] "Ranking index" refers to the ranking or position of an individual within a group, based on their evaluation scores from high to low or from low to high. It is used to indicate the relative order of an individual within the group.

[0297] The "quantile index" refers to the quantile interval or corresponding quantile of an individual after the distribution of group evaluation values ​​is divided into several intervals according to a certain proportion. It is used to represent the relative percentage position of an individual in the overall distribution.

[0298] "Generative artificial intelligence models" refer to artificial intelligence models that are based on machine learning or deep learning techniques, learned from a large amount of sample data, and can automatically generate text, images or other content based on input information.

[0299] "Prompt statements" refer to a sequence of texts or a set of instructions that are constructed by the system and input into the generative artificial intelligence model. They are used to specify the model's task objectives, constraints, and the data content that needs to be referenced.

[0300] "Analysis and processing" refers to the computational process by which generative artificial intelligence models interpret, summarize, or comment on the state, causes, and trends of the evaluated object or group based on prompts and input data.

[0301] "Predictive processing" refers to the computational process by which generative artificial intelligence models estimate and infer future states or outcomes based on historical or current data and prompts.

[0302] "Visualized data" refers to data structures constructed for graphical presentation on display devices, including coordinate axis data, graphic element data, annotation information, and various parameters used for rendering.

[0303] "Display device" refers to an output device that can receive visual data from an information processing device and present information in the form of images or text, including displays, projection devices or other graphic output terminals.

[0304] "User terminal" refers to a device operated by a user and interacting with a server through a communication network, including computing terminals with display and input functions, such as mobile terminals, computer terminals, or other interactive terminals.

[0305] In one embodiment of the present invention, the server, as an information processing device, comprises a processor, main memory, non-volatile memory, a network interface, and an interface for communication with a display device. The server runs an application program on an operating system (e.g., a Unix-like operating system). This application program uses libraries to perform functions such as preprocessing evaluation data, calculating relative pseudo-evaluations, constructing prompt statements, and calling generative artificial intelligence models. The terminal can be a mobile terminal or a computer terminal including a display screen and an input device, through which the user interacts with the server.

[0306] The server installs and invokes data processing and machine learning software in its memory. For example, the server uses a scripting language runtime environment (e.g., the runtime environment of an interpreted programming language) and data analysis libraries (e.g., libraries for tabular data processing, libraries for numerical computation), and machine learning libraries (e.g., libraries containing standardization algorithms and neighborhood search classification algorithms) to perform data standardization and relative pseudo-evaluation modeling. The server also deploys generative artificial intelligence models, which can be deep learning models based on multi-layer neural networks with encoder-decoder or decoder stacked structures, employing multi-head attention mechanisms and feedforward subnetworks, to generate analytical text or suggested content based on prompts.

[0307] The server stores evaluation information tables, group statistics tables, model parameter files, and visualization configuration data in its storage device. Evaluation information may include the evaluated object identifier, the group identifier, the evaluation time, and scores across multiple dimensions (such as performance scores, ability scores, and satisfaction scores). The server uses data structures to organize this information into a table format with rows and columns, and loads it into main memory for processing when needed.

[0308] When preprocessing the evaluation information, the server uses a data analysis library to calculate statistics for each numerical column, such as mean, standard deviation, and median. For missing values, the server employs a mean- or median-based imputation strategy. Values ​​that significantly deviate from the overall distribution are truncated or replaced according to preset statistical thresholds, thereby reducing the impact of extreme values ​​on subsequent calculations. During standardization, the server performs a linear transformation on each feature dimension to ensure that each dimension numerically satisfies the conditions of zero mean and unit variance. Because all evaluation dimensions undergo a uniform scaling transformation, the server can avoid any single dimension dominating distance calculations due to excessively large dimensions when subsequently calculating distances between samples, reducing numerical instability and improving the accuracy and robustness against spurious evaluations.

[0309] When calculating relative pseudo-evaluations, the server employs a neighborhood search classification algorithm. The server uses standardized evaluation data as a set of feature vectors and searches for the nearest neighbors for each evaluated object in the feature space. Based on the labels or score distributions of these neighbors, the server assigns a performance level or calculates the relative position of the overall score for the target object. The server also calculates overall group statistical indicators from the standardized evaluation data of all group components, such as the mean, standard deviation, and percentiles (e.g., 25%, 50%, 75%) for each dimension, and generates difference indicators (the difference between an individual's score and the group mean or median), ranking indicators (ranking or proportional position within the group), and quantile indicators (the quantile interval of an individual's score). These indicators are stored in memory using key-value pairs or structured records for easy retrieval and combination.

[0310] The server, through the aforementioned relative pseudo-evaluation calculation, obtains not only a single absolute score but also a composite index set encompassing multiple statistical relationships. Unlike traditional processing methods based solely on rules or simple ranking, the server automatically captures the similarity between objects by performing neighborhood search in a standardized feature space, utilizing the distance relationships between multi-dimensional features. This allows for a more accurate relative positional judgment that better reflects the overall data distribution without relying on manually set weights. This approach enables the server to maintain high classification stability and scalability on large-scale datasets, ensuring comparability even with changes in group size.

[0311] When constructing prompts for the generative AI model, the server does not simply concatenate natural language templates; instead, it employs structured data mapping rules. The server maps the relative pseudo-evaluation results of the target object and its corresponding difference, ranking, and quantile indices to predefined semantic slots, and generates prompts according to a preset format. For example, the server can generate the following prompts: "Based on Employee A's performance rating data and the evaluation data of all members of the sales department this year, please generate an analysis report explaining Employee A's relative pseudo-evaluation results, including: 1) the difference between Employee A and the department average level; 2) Employee A's strengths and weaknesses in key indicators; 3) suggestions for subsequent incentive and growth management." or: "Based on the relative pseudo-evaluation data of the sales department calculated by the system, please generate a brief Chinese report for employee A that is suitable for reporting to senior management. The report should be within 500 words and should highlight employee A's advantages compared to colleagues at the same level and areas for further improvement." When generating the above prompts, the server associates semantic slots such as "difference from the department average level" and "key indicator strengths and weaknesses" with specific difference indicators and quantile indicators. This allows the generative AI model to explicitly refer to these technical indicators when generating text, thereby improving the interpretability and consistency of the output results.

[0312] In one implementation, the generative AI model used by the server can be a deep neural network containing multiple layers of self-attention and feedforward sublayers. During the model training phase, the server uses a large number of samples containing "input data descriptions (including evaluation metrics and statistical metrics) + target description text" to perform supervised learning on the model. The server uses the cross-entropy loss function as the error function to measure the difference between the model output and the target text, and updates the network weights through the backpropagation algorithm. During training, the server can also employ techniques such as learning rate decay, gradient pruning, regularization, and data augmentation (e.g., changing the wording in the prompts, shuffling some descriptions) to improve the model's generalization ability and robustness across different prompts.

[0313] During the inference phase, the server encodes the constructed prompts into an embedding vector sequence. After feature transformation via a multi-layer attention network, the server generates the analyzed text or suggested content at the decoding end. The server employs a probability distribution selection strategy for the output tokens at each step and can control the diversity and stability of the output using temperature parameters and sampling strategies. Compared to traditional text generation methods based on fixed rule templates, this generation method, based on deep neural networks and learned language patterns, can adaptively adjust the expression while maintaining content consistency, reducing the burden of manually maintaining complex rules.

[0314] When generating visualization data, the server maps relative pseudo-evaluation results, group statistical indicators, and the output of the generative AI model into a unified data structure. For example, the server generates a set of records for each evaluated object, including an identifier, overall score, difference, quantile, performance level, and summary text. The server packages these records into a structure suitable for graphical rendering, such as axis data for bar charts, multi-dimensional coordinate data for radar charts, and explanatory content for text area display. After receiving the visualization data, the terminal uses a graphics rendering library to draw bar charts, line charts, radar charts, or ranking lists on the display screen, while simultaneously displaying explanatory text output by the generative AI model in the text area. Users can intuitively observe the differences between the target object and the group average level and different quantiles through the terminal, and read the explanations and suggestions generated by the system.

[0315] The server achieves multi-level technical improvements within the computer through the aforementioned integrated data standardization, relative pseudo-evaluation calculation, prompt statement construction, and generative AI model invocation process. First, by employing unified standardized processing and statistical indicator calculation, the server avoids numerical biases caused by inconsistent evaluation scales across different times and groups, thereby improving the stability and comparability of the calculation results. Second, by calculating relative pseudo-evaluations through neighborhood search and multi-indicator combinations, the server reduces reliance on manual rules and weight settings, enabling the system to maintain reasonable computational complexity and a low error rate even when processing high-dimensional, large-scale data. Third, by embedding structured statistical indicators into prompt statements and driving the generative AI model, the model can generate more accurate explanatory text within a unified data context, reducing the risk of significant fluctuations in output content due to subtle changes in prompt statements, thus improving the stability and controllability of the natural language generation module within the server.

[0316] In another embodiment, the server can employ different machine learning algorithms to replace the neighborhood search classification algorithm. For example, it can use ranking algorithms based on linear or tree models, clustering algorithms to divide the group into hierarchical levels, or embedding spaces based on metric learning to further optimize the computational efficiency and accuracy of relative pseudo-evaluations. In yet another embodiment, the server can use generative artificial intelligence models with different structures, such as networks employing a hybrid convolutional and attention structure, or sequence-to-sequence models with a separate encoder-decoder architecture. As long as the server can generate explanatory text or suggested content corresponding to the evaluation data based on the prompts, the technical concept of this invention can be applied.

[0317] In some implementations, the server can compress and trim the visualization data based on network bandwidth and terminal processing capabilities. For example, it can send only the data or statistical summaries relevant to the target object to the terminal, thereby reducing communication load and terminal rendering pressure. In this case, the server performs most of the calculations and filtering internally, limiting the terminal's burden to graphics rendering and simple interactions, thus achieving a reasonable allocation of computing resources and improved communication efficiency as a whole system.

[0318] Through the above-described structure, the server improves upon traditional computer processing methods at multiple levels, including data preprocessing, relative pseudo-evaluation modeling, prompt statement construction, and generative artificial intelligence model invocation. This enables the system to automatically generate and visualize evaluation results in large-scale, multi-dimensional data environments with lower computational costs and higher accuracy, rather than simply automating manual analysis processes. Users can more effectively understand and utilize the technical analysis results provided by the system by viewing relative pseudo-evaluation charts and reading explanatory text generated by the generative artificial intelligence model on their terminals.

[0319] use Figure 13 The processing flow is explained.

[0320] Step 1: The server receives the terminal request and parses the parameters.

[0321] Input: Request parameters from the terminal (e.g., department identifier, target object identifier, evaluation period, type of analysis to be output).

[0322] The server reads fields such as department name, evaluated object ID, and time range from the URL parameters or request body of the HTTP request, and validates the field format and value range (e.g., checking if the department exists, whether the object ID conforms to the predetermined rules, and whether the time range is valid).

[0323] Based on the parsing results, the server generates a query configuration data structure in memory (e.g., key-value pairs such as Department=Sales Department, Object ID=Employee A, Year=2024, etc.) as input for subsequent data acquisition and processing.

[0324] Output: Query configuration data used for data querying and calculation.

[0325] Step 2: The server retrieves evaluation information from the data storage device.

[0326] Input: The query configuration data generated in step 1.

[0327] The server constructs query statements based on conditions such as department and time range through database drivers or data access interfaces, and reads the evaluation records of the target group (such as a department) within a specified time period from relational databases or other data storage.

[0328] The server converts the record set returned by the database into a tabular data structure, organizing it by rows corresponding to group components and by columns corresponding to evaluation dimensions (such as performance scores, ability scores, satisfaction scores, etc.).

[0329] The server also marks the row containing the target object so that its relative pseudo-evaluation can be calculated later.

[0330] Output: A raw evaluation data table containing evaluation records of all members of the target group, and identification information for the target object row.

[0331] Step 3: The server performs missing and outlier handling on the evaluation information.

[0332] Input: The original evaluation data table obtained in step 2.

[0333] The server calculates basic statistics (such as mean, standard deviation, and median) for each numerical field, identifies cells containing missing values, and fills in the missing values ​​according to a preset strategy (such as using the mean or median).

[0334] The server identifies outlier values ​​that significantly deviate from the overall distribution based on statistical rules (e.g., absolute deviation exceeding a certain number of standard deviations) and performs truncation, replacement, or record removal operations on these values.

[0335] After completing missing value imputation and outlier handling, the server generates a cleaned evaluation data table, ensuring that each record is a valid value in each evaluation dimension.

[0336] Output: A cleaned evaluation data table with missing values ​​filled and outliers handled.

[0337] Step 4: The server performs standardization processing on the cleaned evaluation data.

[0338] Input: The post-cleaning evaluation data table obtained in step 3.

[0339] The server calculates the mean and standard deviation for each evaluation dimension (such as performance score, ability score, satisfaction score, etc.) and performs a linear transformation on each data point to convert each dimension into a standardized value with zero mean and unit variance.

[0340] The server reorganizes the standardized results of each dimension into a feature matrix. In the feature matrix, each row corresponds to a group constituent element, and each column corresponds to a standardized feature.

[0341] The server retains the mean and standard deviation parameters during the standardization process so that new data can be transformed to the same scale later.

[0342] Output: Standardized evaluation data matrix, and corresponding standardized parameters (mean and standard deviation).

[0343] Step 5: Overall statistical indicators of the server computing group.

[0344] Input: The standardized evaluation data matrix obtained in step 4.

[0345] Based on standardized data, the server calculates overall statistical indicators for the group for each evaluation dimension, such as the overall mean, overall variance, maximum value, minimum value, and several quantiles (e.g., 25%, 50%, 75%).

[0346] The server calculates a comprehensive score for each group component (e.g., by weighted summation of multiple standardized dimensions or other combination functions), and sorts all objects accordingly to obtain ranking information and quantile positions.

[0347] The server stores the above-mentioned group statistical indicators and the ranking, percentile, and other results corresponding to each object as structured records, which serve as the basis for the subsequent generation of relative pseudo-evaluations and prompt statements.

[0348] Output: A set of overall statistical indicators for the group and a statistical results table containing the comprehensive score, ranking, and percentile of each object.

[0349] Step 6: The server calculates a relative pseudo-evaluation based on neighborhood search.

[0350] Input: The standardized evaluation data matrix from step 4 and the statistical results table from step 5.

[0351] In the feature space, the server uses the standardized feature vectors of each group's constituent elements as sample points and calculates the similarity between samples using distance metrics (such as Euclidean distance or cosine similarity).

[0352] For each target object (including at least the specified target object), the server finds several nearest neighbor samples, calculates the comprehensive score distribution or predefined grade distribution of these neighbors, and classifies or judges the relative position of the target object based on the distribution to obtain a relative pseudo-evaluation result.

[0353] The server combines relative pseudo-evaluations with overall group statistical indicators to form relative indicators used to describe "how much above the average" or "what percent of the group is in the top percentile".

[0354] Output: A set of relative pseudo-evaluation results for each target object, including indicators such as relative rank, difference from the group mean, ranking, and quantile.

[0355] Step 7: The server constructs prompts for use by generative artificial intelligence models.

[0356] Input: The set of relative pseudo-evaluation results generated in step 6 and the set of statistical indicators from step 5.

[0357] The server extracts information such as the target object's comprehensive score, the difference from the group average, quantile, and ranking from the set of relative pseudo-evaluation results, and maps these values ​​into natural language expressions according to pre-set semantic templates and rules.

[0358] The server assembles the prompt statement according to a predetermined structure, including task description, data background, and key output points. For example, the server generates the following prompt statement: "Based on Employee A's performance rating data and the evaluation data of all members of the sales department this year, please generate an analysis report explaining Employee A's relative pseudo-evaluation results, including: 1) the difference between Employee A and the department average level; 2) Employee A's strengths and weaknesses in key indicators; 3) suggestions for subsequent incentive and growth management." The server can also generate brief prompts for high-level reporting, such as: "Based on the relative pseudo-evaluation data of the sales department calculated by the system, please generate a brief Chinese report for employee A that is suitable for reporting to senior management. The report should be within 500 words and should highlight employee A's advantages compared to colleagues at the same level and areas for further improvement." Output: Prompt text corresponding one-to-one with relative pseudo-evaluations and statistical indicators.

[0359] Step 8: The server invokes a generative artificial intelligence model to generate analytical text or suggested content.

[0360] Input: The prompt text generated in step 7.

[0361] The server converts the prompt statement into the input format required by the model (such as a tokenized sequence after word segmentation or sub-word encoding) and sends it to the model computation module through the inference interface of the generative artificial intelligence model.

[0362] The server uses a multi-layer neural network inside the model to encode and decode the prompts, extracts semantic features through an attention mechanism, and generates analysis text or suggestions word by word.

[0363] The server receives the text output generated by the model and performs basic processing on the output, such as length trimming, filtering illegal characters, or checking for sensitive information.

[0364] Output: Analytical text, explanatory content, or management suggestion text tailored to the target object.

[0365] Step 9: The server generates visualized data and sends it to the terminal.

[0366] Input: The relative pseudo-evaluation result set obtained in step 6, the population statistical indicators obtained in step 5, and the analysis text obtained in step 8.

[0367] The server assembles the target object's overall score, departmental average score, percentile, and other numerical values ​​into chart data structures, such as category labels and corresponding values ​​for bar charts, and multi-dimensional coordinate values ​​for radar charts.

[0368] The server will analyze the correlation between the text and the relative pseudo-evaluation results to form a unified visualization data package, which includes both the numerical data required for graphing and explanatory content for displaying text areas.

[0369] The server sends the visualization data packet to the terminal as a response message via the network interface.

[0370] Output: A terminal-oriented visualization data package containing chart data and explanatory text.

[0371] Step 10: The terminal renders and displays the evaluation results and analysis content.

[0372] Input: The visual data packet sent by the server in step 9.

[0373] The terminal parses chart configuration data and text content from data packets and calls the local graphics rendering component to draw visual elements such as bar charts, line charts, radar charts, or ranking lists on the display screen.

[0374] The terminal displays analysis descriptions generated by a generative artificial intelligence model in the text area near the charts, allowing users to see both numerical comparisons and textual interpretations on the same interface.

[0375] Based on the user interface's interactive design, the terminal allows users to switch between different objects or time periods and resend requests to the server when the user interacts with the device, thereby enabling dynamic updates and display of the evaluation results.

[0376] Output: A visual interface displayed on the terminal screen, including relative pseudo-evaluation charts and corresponding analysis text.

[0377] Step 11: Users can view and utilize the system output on the terminal.

[0378] Input: The charts and text information displayed on the terminal in step 10.

[0379] Users can identify the relative position and performance differences of the target object relative to the group by observing the charts; and understand the background of the differences and possible directions for improvement by reading the explanatory text.

[0380] Based on the technical analysis results provided by the system, users can adjust goal settings, develop training programs, or configure incentive measures in the external management system, but these external operations do not change the internal technical processing flow of the server.

[0381] Output: The user's understanding of the system output and subsequent decision-making behavior (as an external effect of the system, it will not be fed back as new program output).

[0382] 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".

[0383] In existing technologies, personnel management systems, performance evaluation systems, target management systems, work monitoring systems, and machine learning systems are mostly independent applications with varying data formats and inconsistent update cycles. This results in servers being able to process data on a single dimension (e.g., performance or equipment utilization alone) during comprehensive analysis, making it difficult to create a unified performance view across personnel and work objects. Furthermore, traditional systems typically treat users' emotional states as unstructured or noisy information, and servers do not utilize emotional information to adjust algorithm parameters or output results during data processing, leading to the following technical problems: (1) From the perspective of computer system architecture, the data flow on the server side is fragmented. Personnel-related data, business process data and emotional data are scattered in different storage and different service interfaces. The lack of a unified data model and processing pipeline makes it impossible for the server to synchronously process and jointly model multi-source heterogeneous data in the same computing path, which reduces data processing efficiency and increases the complexity of model calling and scheduling.

[0384] (2) From the perspective of model invocation and inference process, when existing systems invoke generative artificial intelligence models, most of them only generate fixed-format input based on static business data. The prompt statements are simple in design and do not structure the performance indicators, statistical results and sentiment evaluation indicators calculated by the machine side into the prompt statements. As a result, the generative artificial intelligence model cannot make full use of the existing calculation results of the server and the analysis quality depends on the model to "find structure from data", which increases the inference burden and uncertainty.

[0385] (3) From the perspective of performance evaluation and decision support, traditional solutions often rely on batch-processed statistical reports, which can only provide a lagging, one-dimensional evaluation of the performance of personnel and work objects. It is difficult to perform relative pseudo-evaluation calculations and dynamic rankings across members and devices. In particular, it is impossible to comprehensively quantify emotional state and performance indicators under the same computing framework. Therefore, it is difficult to provide fine-grained input features for subsequent task priority adjustment, resource scheduling and load balancing algorithms of the server.

[0386] (4) From the perspective of visualization and human-computer interaction, the server and the terminal mostly transmit static reports or simple alarm information. The terminal cannot obtain a high-dimensional visualization data structure based on a unified comprehensive dataset and generative artificial intelligence model results. The server also has difficulty optimizing the logic of prompt statement generation and model calling strategy based on user feedback in a closed loop, resulting in insufficient adaptive capability of the overall system at the human-computer collaboration level.

[0387] Therefore, how can we design a unified data processing and model linkage architecture on the server side, enabling the server to: - Integrate personnel-related information, business-related information, and sentiment-related information into a single data structure for comprehensive calculation; - Automatically generate structured and semantically rich prompts based on comprehensive datasets and performance metrics to drive generative artificial intelligence models to perform high-quality analysis and prediction; - Implement relative pseudo-evaluation calculation, ranking, and visualization generation within the same processing chain; - The generated results are fed back to the terminal and user in both machine-readable and visual formats, thereby optimizing the server's data processing and model calling process. This has become a computer technology problem that urgently needs to be solved.

[0388] 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.

[0389] In this invention, the server includes a personnel-related information processing unit, a business-related information processing unit, an emotion-related information processing unit, a comprehensive data generation unit, a prompt statement generation unit, a generative artificial intelligence model linkage unit, a visualization processing unit, and an information providing unit. This creates a unified processing pipeline within the server, connecting multi-source data acquisition, feature calculation, prompt statement construction, generative artificial intelligence model inference, and visualization output. This enables the computer system to: integrate personnel information, job object information, and emotion evaluation indicators within a unified data structure; automatically generate prompt statements for specific analysis tasks based on the comprehensive dataset; and inject structured calculation results into the generative artificial intelligence model input, reducing the model's burden of parsing raw unstructured data and improving the relevance and stability of inference results. Simultaneously, the server performs relative pseudo-evaluation calculation and ranking; the visualization processing unit transforms performance indicators, relative positions, and emotion-performance correlations into display data, which is then sent to the terminal via the information providing unit. This achieves integrated decision support for personnel management and job object management, and further optimizes the server's prompt statement generation strategy and model linkage strategy through user feedback, thereby substantially improving the computer system's processing performance and intelligence level in multi-source data fusion analysis and human-machine collaborative decision-making.

[0390] A "system" refers to a collection of integrated computer devices consisting of multiple information processing units, storage units, and communication units connected through a communication network, used to perform data acquisition, processing, analysis, visualization, and information provision.

[0391] The "Personnel-related Information Processing Unit" refers to a data processing module used to obtain personnel-related attribute information, evaluation information, target information, and progress information from data sources, and to organize, transform, and store such information.

[0392] The "Business-Related Information Processing Unit" refers to a data processing module used to obtain operational information, production information, fault information, and operational status information related to the work object from the data source, and to clean, classify, and structure such information.

[0393] The “emotion-related information processing unit” refers to a data processing functional module used to acquire user emotional state information, analyze, identify and quantify the information to generate emotional evaluation indicators.

[0394] The "Comprehensive Data Generation Unit" refers to a data processing module that integrates personnel-related information, business-related information, and sentiment evaluation indicators into a unified data structure, and generates a comprehensive dataset and computational performance indicators based on the integrated data.

[0395] A “comprehensive dataset” refers to a unified structured data collection that includes various types of data such as personnel-related information, business-related information, and sentiment evaluation indicators, and is used for subsequent analysis, calculation, and model input.

[0396] "Performance metrics" refer to quantitative indicators calculated based on a comprehensive dataset to represent the performance level of personnel or work objects under certain time or conditions, including but not limited to efficiency values, output levels, relative rankings, and trend values.

[0397] The "prompt statement generation unit" refers to a functional module that automatically generates natural language or structured descriptive text to instruct generative artificial intelligence models to perform specific analysis or prediction tasks based on a comprehensive dataset and performance metrics.

[0398] "Prompt statements" refer to the instruction text or combination of text and structured information input into a generative artificial intelligence model to explicitly specify the analysis objectives, constraints, data summaries, and expected output formats.

[0399] The “Generative Artificial Intelligence Model Linkage Unit” refers to the interface and control function module used to send prompts and related data to the generative artificial intelligence model, receive the output results of the model, and parse and process them for subsequent use.

[0400] "Generative AI models" refer to AI models that can automatically generate text, metrics, or structured results based on input prompts and related data, including but not limited to natural language generation models that use deep learning algorithms.

[0401] "Pseudo-evaluation" refers to a presumptive quantitative indicator that is not based on manual scoring results, but rather on data analysis and model predictions, and is used to approximate the evaluation level of personnel or work objects.

[0402] "Relative pseudo-evaluation" refers to a presumptive evaluation index that represents the relative position or ranking relationship by comparing and calculating the pseudo-evaluations of multiple members or multiple work objects.

[0403] "Suggestion information" refers to natural language or structured suggestion data output by generative artificial intelligence models or data processing units based on comprehensive datasets and performance indicators, used to guide business organization, efficiency improvement, or task adjustment.

[0404] The "visualization processing unit" refers to a graphics processing module that uses performance indicators, pseudo-evaluations, relative pseudo-evaluations, and recommendation information to generate charts, graphs, or other visually presentable data for intuitive display on a display device.

[0405] "Display data" refers to formatted and graphical data generated by a visualization processing unit for presentation on a display device, including but not limited to line charts, bar charts, ranking lists, and text descriptions.

[0406] "Information providing unit" refers to a communication and output control function module used to send display data and related suggestion information to terminals or other output devices via a network, so that users and managers can browse, operate and use such information.

[0407] "User" refers to an individual user who uses the system functions through a terminal, receives evaluation results and suggestions, and adjusts their own tasks or behaviors accordingly.

[0408] "Manager" refers to a user with management authority who can view the comprehensive evaluation results and suggestions of multiple users or multiple work objects through the system, in order to implement resource allocation, performance management and strategy adjustment.

[0409] "Work object" refers to the object that is monitored and evaluated in the course of business operations, including but not limited to mechanical equipment, automated devices, or other entities that can be quantified and evaluated to perform specific work tasks.

[0410] "Progress information" refers to quantitative or semi-quantitative data used to represent the degree of completion, time nodes, and remaining workload of personnel or work objects in a given goal or plan.

[0411] "Ranking information" refers to data indicating the ranking or order of multiple personnel or work objects after ranking them based on performance indicators or relative pseudo-evaluations.

[0412] "Relative position information" refers to data used to represent the positional relationship of a person or work object relative to other persons or work objects in an evaluation dimension, including percentiles, grade intervals or grouping results.

[0413] "Task priority adjustment" refers to the process and result of resetting the execution order or resource allocation weights of multiple tasks based on performance indicators, emotional evaluation indicators and suggestion information.

[0414] "Motivation improvement" refers to the effect of enhancing the subjective enthusiasm of users by continuously providing pseudo-evaluations, trend information and personalized suggestion information to guide users to improve their behaviors and increase participation.

[0415] "Growth management" refers to the process in which managers track, evaluate and develop training plans for the long-term development of users based on the historical data, trend analysis and emotion-related indicators output by the system.

[0416] In various embodiments of the present invention, a server, terminals and users cooperatively form a computer system for comprehensively processing personnel-related information, service-related information and emotion-related information, and linking with a generative artificial intelligence model. Through specific data structures, specific feature extraction methods, specific prompt sentence construction algorithms, and specific model calling and visualization processes, the server realizes efficient fusion and intelligent analysis of multi-source heterogeneous data, thereby improving traditional computer technology in terms of computing speed, analysis accuracy, data management and communication load.

[0417] 1. Hardware and basic software structure In one embodiment, the server comprises a multi-core general-purpose processor, a main memory, a high-speed cache, a persistent storage device and a network interface device. The server runs a general-purpose operating system, such as a Unix-like operating system or a server operating system. The server installs a database management system, such as a relational database system, in the storage device, which is used to store personnel information tables, business data tables, emotion data tables, comprehensive data set tables, pseudo-evaluation result tables, etc. The server installs a script operating environment and scientific computing libraries at the application layer, for example: The server uses a programming language to run application logic; The server uses data analysis libraries (such as data frame libraries and numerical computing libraries) to perform batch data processing; The server uses visualization frameworks (such as graphics drawing libraries and Web visualization frameworks) to generate dynamic charts and dashboards; The server uses deep learning frameworks (such as general tensor computing frameworks) to load and infer neural network models; The server communicates with external generative artificial intelligence model services through the HTTP or HTTPS protocol.

[0418] In one implementation, a terminal is a mobile computing device, such as a smartphone or tablet. The terminal includes a processor, memory, display screen, camera, microphone, and wireless communication module, runs a mobile operating system, and has a browser or dedicated applications installed. The terminal interacts with a server via a network to collect user input and display visualizations and suggestions generated by the server.

[0419] Users interact with the server through a terminal in the system, inputting necessary business and emotional information, and performing real-world adjustments to equipment configuration or personnel tasks based on the server's output.

[0420] II. Structured processing of personnel-related and business-related information by the server In one implementation, the server stores personnel-related information and business-related information separately as multidimensional table structures. The server uses a unified field naming and data type specification for each type of information. For example: The server sets fields in the "Personnel Information Table": Personnel Identifier, Organization, Job Category, etc. The server sets fields in the "Evaluation Information Table": Personnel Identifier, Time, Evaluation Score, Evaluation Dimension, etc. The server sets fields in the "Target Information Table": Personnel Identifier, Target Period, Target Type, Target Value, etc. The server sets fields in the "Progress Information Table": Personnel Identifier, Task Identifier, Completion Percentage, Update Time, etc. The server sets fields in the "Job Object Information Table": job object identifier field, type field, location field, etc. The server sets the following fields in the "Operation Information Table": Operation Object Identifier, Time Period, and Operation Duration. The server sets the following fields in the "Production Information Table": Job Object Identifier, Output Quantity, and Defect Quantity. The server sets fields in the "Fault Information Table" such as: job object identifier field, number of faults field, and downtime field.

[0421] The server uses a data analysis library to read the data from the above tables into a data frame structure, imputing missing values ​​(e.g., using the median or grouped mean) and filtering outliers using statistical methods (e.g., using a deviation multiple threshold). The server joins multiple tables by personnel and task identifiers through indexing and grouping operations, constructing a comprehensive data record containing multidimensional features. The server calculates performance-related characteristics from this comprehensive data, such as: The server calculates average task completion rate, target achievement rate, and average customer feedback for personnel. The server calculates unit time output, failure rate, and overall efficiency for the task.

[0422] This structured processing approach allows servers to store multi-source data using unified key-value and field types, facilitating subsequent numerical calculations and model feature construction, reducing the overhead of data format conversion, and thus improving overall processing speed and data management consistency.

[0423] III. Feature Extraction and Quantification of Emotion-Related Information by the Server In one implementation, the server collects raw emotional data from user text, voice, or image input via a terminal. The terminal then sends a request containing text content or multimedia data to the server.

[0424] For text-based sentiment data, in one implementation, the server uses a natural language processing library to segment the text, filter stop words, and map the text into a vector representation using pre-trained word vectors or embedding vectors. Based on this, the server employs a classification model, such as a sentiment classification model based on a bidirectional recurrent network with an attention mechanism or a multi-layer self-attention network. During inference, the server uses pre-trained model weights to output the sentiment category (e.g., positive, neutral, negative) and the sentiment intensity score (e.g., a continuous value from 0 to 1). The server uses the sentiment category and intensity value as "sentiment evaluation indicators" and writes them into the "sentiment data table."

[0425] For image or speech-based emotion data, the server uses convolutional neural networks to extract facial expression features or a combination of convolutional and recurrent structures to extract speech features in one implementation. The server loads a pre-trained network model into a deep learning framework. For example, for image expression recognition, the server uses a network consisting of multiple convolutional, pooling, and fully connected layers to generate feature vectors for facial regions, and then outputs various emotion probabilities through Softmax. For speech emotion recognition, the server extracts spectral or Mel-frequency coefficient features and inputs them into a convolutional-recurrent hybrid network to obtain the emotion classification result. The server quantifies the predicted emotion labels and confidence scores into emotion evaluation metrics.

[0426] The server transforms previously unstructured multimedia sentiment information into numerical indicators, uniformly incorporating them into a comprehensive dataset. This allows sentiment states to participate in performance evaluation calculations and prompt generation as standardized features in subsequent computations. This quantification and structuring process reduces dimensional redundancy and noise propagation caused by directly inputting text or multimedia into the model, improving computational efficiency and analytical accuracy.

[0427] IV. Server Calculation of Comprehensive Data Set and Performance Metrics In one implementation, the server, through a comprehensive data generation unit, performs multiple joins between personnel-related tables, business-related tables, and sentiment data tables based on personnel identifiers, task object identifiers, and the time dimension to generate a comprehensive dataset data structure. This comprehensive dataset contains multiple feature vector fields, such as: The server creates a record for each person in each time period. The record includes: performance score, task completion rate, goal achievement, emotional intensity, main emotion category, and trend indicators of recent periods. The server establishes a record for each job object in each time period. The record includes: total operating time, total output, defect rate, failure frequency, and average emotional intensity of the personnel involved in the operation.

[0428] The server uses a numerical library to perform batch calculations on the comprehensive dataset, calculating performance metrics, including but not limited to: weighted efficiency scores, trend scores, and outlier scores. In some implementations, the server employs methods such as standardization and principal component analysis to reduce the dimensionality of high-dimensional features, thereby reducing the dimensionality of subsequent model inputs, improving computational efficiency, and mitigating the risk of overfitting.

[0429] V. Server-side generation of prompt statements and its interaction with the generative artificial intelligence model In one implementation, the server uses a comprehensive dataset and performance metrics as input to generate prompts. The server constructs the prompts using a combination of template-driven and rule-selective methods. The server pre-stores multiple prompt templates, each corresponding to a specific analysis task. For example: For tasks comparing the efficiency of two job objects, the server uses a similar template: "Robot A produced X items in the last hour and operated for Y hours; Robot B produced M items in the last hour and operated for N hours. Please compare their efficiencies and identify the more efficient robot, while also providing specific suggestions for improving the less efficient robot." For tasks involving employee performance and sentiment analysis, the server uses a similar template: "Below is employee A's performance data for the past 6 months (sales, project completion rate, customer feedback rating, etc.), as well as sentiment analysis results (overall mood bias, stress level). Please generate a pseudo-evaluation score from 0 to 100 for each month and provide three suggestions for alleviating stress and maintaining or improving performance." For tasks requiring priority adjustment, the server uses a similar template: "The following is a list of tasks that the user has not yet completed, along with their urgency and estimated time. The current sentiment analysis result is 'high stress.' Please adjust the task priorities, temporarily postponing high-intensity tasks, and provide the new task order and the reasons for the adjustment." When generating prompts, the server fills placeholder variables in the template with numerical features from the comprehensive dataset, forming a natural language description containing specific numbers and background information. This structured prompt generation method allows generative AI models to "see" the server's completed calculations when receiving requests, rather than inferring structure from unstructured data. This reduces the model's internal reasoning burden and improves the consistency and relevance of the output with existing system data.

[0430] The server sends the prompt and necessary simplified data summary to the generative AI model service via a network interface. In one implementation, the generative AI model is a multi-layered self-attention sequence-to-sequence generation model with an encoder and decoder structure. The server specifies the maximum output length, temperature parameters, and sampling method during invocation to control the determinism and diversity of the generated text.

[0431] In another implementation, the server deploys a generative model based on a deep learning framework locally and loads pre-trained weights. After receiving prompts, the server encodes them into vector sequences and inputs them into the encoder. The decoder then gradually generates output labels. The parameters obtained by optimizing the cross-entropy loss function ensure that the model has learned reasonable language and reasoning capabilities during the training phase.

[0432] VI. Server-side parsing and visualization of the generated results After receiving the text or structured output returned by the generative artificial intelligence model, the server parses it. Based on preset labeling rules or simple semantic segmentation methods, the server divides the results into parts such as "pseudo-evaluation score segments," "explanation of reasons segments," and "suggestion segments." The server parses the pseudo-evaluation scores into numerical values ​​and stores them in the results table along with the original performance indicators.

[0433] The server utilizes a visualization library to plot performance metrics, pseudo-evaluations, relative pseudo-evaluations, and sentiment indicators. For example, the server can create time-series charts for personnel, overlaying real performance curves and pseudo-evaluation curves, and using color coding to represent changes in sentiment status; it can also create ranking bar charts for task objects, labeling efficiency levels and relative differences; and it can create scatter plots to represent the correlation between sentiment intensity and performance indicators. The server encapsulates these charts into a front-end renderable data format using a web framework and provides access interfaces.

[0434] By employing this structured parsing and graphical processing method, the server eliminates the need for extensive local computation when presenting data on the terminal, thereby reducing the terminal's processing load and power consumption. Simultaneously, the server pre-compiles coordinate transformations, aggregation calculations, and layout calculations related to chart rendering, improving overall rendering speed and reducing the amount of data transmitted over the network.

[0435] VII. Terminal Display and User Interaction In one implementation, the terminal accesses a dashboard page provided by the server through a browser; in another implementation, it retrieves and displays data by calling the server interface through a dedicated application. The terminal renders the visualized data returned by the server as line charts, bar charts, leaderboards, and suggestion lists. Users can intuitively view their own or their subordinates' performance trends, changes in pseudo-evaluations, and their relative positions with other members on the terminal.

[0436] The terminal also provides users with a form interface for inputting subjective emotional descriptions or selecting the current emotion tag. When necessary, the terminal activates the camera and microphone for image or audio capture to support more refined emotion recognition. After receiving task priority adjustment suggestions from the server, the terminal prompts the user through the interactive interface whether they accept the adjustment and feeds back the selection result to the server so that the server can use user feedback as an iterative signal in subsequent analysis.

[0437] VIII. Technical Effects and Causal Relationships Through the specific data structure design, feature quantization method, prompt statement generation, and model linkage process described above, the server improves upon traditional computer technology in the following aspects: The server manages data sources that were originally scattered across different systems in a single data structure through a unified comprehensive dataset and standardized fields, thereby reducing cross-database connections and redundant transmissions between multiple databases, shortening data processing paths, and improving processing speed. By pre-calculating performance and sentiment metrics locally and embedding these results into prompts, the server eliminates the need for generative AI models to re-infer feature structures from raw unstructured data, reducing the computational load of model inference and improving the accuracy and stability of text generation results. The server performs most of the numerical calculations internally through relative pseudo-evaluation and ranking algorithms, compressing the visual data that only needs to be displayed on the terminal into a smaller size, thereby reducing the communication load. By incorporating sentiment indicators into performance calculations and prompt generation rules, the server enables the system to adopt a non-traditional human subjective judgment mode when calculating evaluations and adjusting task priorities. Instead, it performs machine-specific weight adjustments and rule combinations based on multi-dimensional features and statistical results, which helps reduce decision-making errors caused by a single indicator. The server employs a multi-layer neural network structure under the deep learning framework to achieve feature extraction and emotion recognition of multimodal data such as text, images, and speech. It optimizes model parameters using general methods such as backpropagation and gradient descent, and performs fast vector operations with fixed weights during the inference phase, thereby providing high throughput and low latency in large-scale data processing scenarios.

[0438] IX. Optional Implementation Forms and Variations In some implementations, the server can employ only text sentiment recognition, without enabling image or audio data processing, to simplify the computation process and adapt to scenarios with limited terminal hardware capabilities. In another implementation, the server can automatically assess the quality of the output of the generative AI model, regenerating prompts for outputs that do not meet preset rules and then calling the generative AI model again, thereby improving the overall output quality.

[0439] In some implementations, the server can cache the generated prompts and corresponding model outputs as samples for subsequent fine-tuning of the locally generated model, making the model more suitable for specific enterprises or scenarios. In other implementations, the server can also use different neural network structures, such as using convolutional networks to process time-series numerical features and graph neural networks to process organizational structure relationship data, in order to achieve specialized optimization for different types of inputs.

[0440] Through the above-mentioned various implementation forms and replaceable configurations, the server can flexibly select appropriate model structures and parameter settings according to hardware resources, data types and business scale, thereby maintaining system versatility while further improving computing efficiency, analysis accuracy and system stability, and realizing computer technology improvements for multi-source information fusion analysis and human-machine collaborative decision-making.

[0441] use Figure 14 The processing flow is explained.

[0442] Step 1: The server retrieves basic data from the data source.

[0443] Inputs: Personnel information table, evaluation information table, target information table, and progress information table from the personnel database; work object information table, utilization information table, production information table, and fault information table from the business database; and raw emotional data (text, image, or audio) uploaded from the terminal.

[0444] The server performs specific data processing and calculations: it reads raw records from multiple tables through database query statements and loads the results into the data frame structure of the data analysis library; it performs uniform format conversion on the time field, performs type validation on personnel identifiers and work object identifiers, and marks missing fields; the server also receives HTTP requests from terminals and caches the attached emotional text, image or audio data in the temporary storage area.

[0445] Output: A set of raw data that has undergone preliminary verification and format standardization, including structured table data and raw sentiment data to be parsed.

[0446] Step 2: The server cleans and standardizes personnel-related and business-related information.

[0447] Input: Data from the personnel information table, evaluation information table, target information table, progress information table, work object information table, operation information table, production information table, and fault information table output from step 1.

[0448] The server performs the following specific data processing and calculations: It uses a data analysis library to detect missing values ​​in each field and fills in key numerical fields using grouped averages or medians; it uses statistical methods to calculate the mean and standard deviation of each field, identifies outliers exceeding a preset multiple of the standard deviation, and marks or removes them; it performs normalization or standardization on numerical fields (e.g., subtracting the mean and then dividing by the standard deviation) and encodes categorical fields; and it joins and merges multiple tables based on personnel and work object identifiers to form intermediate data frames segmented by time.

[0449] Output: Cleaned and standardized intermediate datasets of personnel-related data and business-related data.

[0450] Step 3: The server extracts emotional features from the raw emotional data and quantifies them into emotional evaluation indicators.

[0451] Input: Emotional text data, emotional image data, or emotional audio data output from step 1.

[0452] The server performs the following specific data processing and calculations: For text data, the server uses word segmentation and vectorization tools to convert the text into vector features, and calls a pre-trained sentiment classification model (based on a deep learning framework) for forward inference to obtain sentiment category and sentiment intensity score; For image data, the server uses an image processing library to extract face regions, inputs the image into a convolutional neural network model, outputs the probability distribution of multiple sentiment categories, and selects the category with the highest probability and its probability value; For audio data, the server first extracts spectrum or sound features, and then inputs them into the sentiment recognition network to obtain sentiment labels and confidence scores.

[0453] Output: A table of emotional evaluation metrics including personnel identifiers, timestamps, emotional categories, and emotional intensity scores.

[0454] Step 4: The server generates a comprehensive dataset and calculates performance metrics.

[0455] Input: The intermediate datasets related to personnel and business, output from step 2; and the sentiment evaluation index data table, output from step 3.

[0456] The server performs the following specific data processing and calculations: Based on personnel identification and timestamps, the server connects the sentiment evaluation index data table with the personnel-related dataset, merging sentiment intensity and performance indicators within the same time window into a single record; Based on the task object identification and time period, the server aggregates and calculates the average value of sentiment features related to specific task objects in the sentiment data; The server calculates performance indicators for each record in the comprehensive dataset, including but not limited to unit time efficiency, target achievement rate, trend change rate, etc., and adds these performance values ​​as new fields.

[0457] Output: A comprehensive dataset containing personnel characteristics, business characteristics, sentiment evaluation metrics, and performance metrics.

[0458] Step 5: The server calculates relative pseudo-evaluations and ranking information.

[0459] Input: The comprehensive dataset output from step 4.

[0460] The server performs the following specific data processing and calculations: The server calculates a pseudo-evaluation score for each object using performance indicators and sentiment evaluation indicators according to preset formulas; the server groups and sorts the pseudo-evaluation scores, and calculates the ranking, percentile, and deviation from the average value of each object within the group; the server generates relative pseudo-evaluation tables and ranking tables in the dimensions of personnel and work objects, respectively.

[0461] Output: A table of relative pseudo-evaluation results containing fields such as pseudo-evaluation score, relative ranking, and percentile.

[0462] Step 6: The server generates prompts for invoking generative artificial intelligence models.

[0463] Input: The comprehensive dataset output from step 4, the relative pseudo-evaluation result data table output from step 5, and the preset prompt statement template.

[0464] The server performs specific data processing and calculations: The server extracts key features from the comprehensive dataset (such as an employee's performance data and emotional changes over the past few months, the efficiency and failure rate of a certain task object, etc.), and reads the relative position data of the corresponding object from the relative pseudo-evaluation result table; The server inserts these values ​​and labels into a preset template to construct a natural language description and form a complete prompt statement; The server selects different templates according to different task types, such as comparing equipment efficiency, generating monthly pseudo-evaluations for employees, and adjusting task priorities.

[0465] Output: A prompt text containing specific numerical values ​​and contextual information, along with the corresponding target object identifier.

[0466] Step 7: The server invokes the generative artificial intelligence model and obtains the analysis results.

[0467] Input: The prompt text output from step 6 and any necessary data summary.

[0468] The server performs the following specific data processing and calculations: The server sends the prompt statement to the generative artificial intelligence model service through the network interface and sets the model parameters (such as maximum output length and generation temperature); the server receives the generated text response, parses it into structured information, and separates the pseudo-evaluation suggestions, explanations of reasons, and improvement suggestions; the server performs format validation on the pseudo-evaluation values ​​in the model output and performs simple cleanup on the suggestion text.

[0469] Output: Structured model analysis results, including supplementary pseudo-evaluation scores, explanation text of reasons, and suggestion information text.

[0470] Step 8: The server integrates the model results with the local calculation results and generates visualized data.

[0471] Inputs: Performance metrics output from step 4, relative pseudo-evaluation results output from step 5, and model analysis results output from step 7.

[0472] The server performs the following specific data processing and calculations: The server compares the pseudo-evaluations output by the model with the pseudo-evaluations calculated locally. If the difference exceeds a preset threshold, it is recorded as an abnormal reference. The server constructs a data structure for plotting, organizing time series performance indicators, pseudo-evaluation scores, and sentiment intensity into multi-dimensional arrays. The server uses a visualization library to calculate chart coordinates, color codes, and legend information to generate the data formats required for line charts, bar charts, and scatter plots.

[0473] Output: A visual dataset for front-end rendering and readable analytical summary text.

[0474] Step 9: The terminal receives visual data and displays the interface.

[0475] Input: The visualized data set and analysis summary text output from step 8.

[0476] The specific data processing and calculation performed by the terminal: The terminal obtains visual data from the server through network requests, maps the data to chart components, and draws line charts, bar charts, and ranking lists; The terminal displays the pseudo-evaluation change trend, relative ranking, and model-generated suggestion information on the interface; The terminal scales and adjusts the layout of the charts according to the screen size and resolution.

[0477] Output: A graphical interface and text suggestions displayed on the terminal screen for the user to view.

[0478] Step 10: Users can input feedback or adjustment instructions based on the displayed results.

[0479] Input: The charts and suggestions displayed in the terminal in step 9.

[0480] The specific operations performed by the user are as follows: The user views the pseudo-evaluations and suggestions for themselves or the managed objects on the terminal interface, and decides whether to accept the task priority adjustment, modify the target, or adjust the job object configuration based on the suggestions; the user clicks the operation buttons such as "Accept", "Reject" or "Save after modification" on the interface, or enters supplementary description text.

[0481] Output: User feedback data and adjustment instructions, which are sent back to the server via the terminal to update the comprehensive dataset and parameter settings used in subsequent analysis.

[0482] 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.

[0483] 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 partially or entirely performed by AI, but is not limited to this example. Furthermore, the processing performed by 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 AI including the generation AI.

[0484] 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.

[0485] 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.

[0486] 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.

[0487] Second Implementation Method Figure 3 An example of the configuration of the data processing system 210 according to the second embodiment is shown.

[0488] 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.

[0489] 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).

[0490] 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.

[0491] 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.

[0492] 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 capture 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).

[0493] 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.

[0494] 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.

[0495] 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.

[0496] 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).

[0497] 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.

[0498] 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".

[0499] 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.

[0500] 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.

[0501] 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.

[0502] 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.

[0503] 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.

[0504] 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 partially or entirely performed by AI, but is not limited to this example. Furthermore, the processing performed by 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 AI including the generation AI.

[0505] 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.

[0506] 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.

[0507] 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.

[0508] Third Implementation Method Figure 5 An example of the configuration of the data processing system 310 according to the third embodiment is shown.

[0509] 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.

[0510] 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).

[0511] 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.

[0512] 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.

[0513] 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 capture 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).

[0514] 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.

[0515] 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.

[0516] 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.

[0517] 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.

[0518] 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.

[0519] 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".

[0520] 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.

[0521] 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.

[0522] 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.

[0523] 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.

[0524] 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.

[0525] 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 partially or entirely performed by AI, but is not limited to this example. Furthermore, the processing performed by 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 AI including the generation AI.

[0526] 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.

[0527] For example, the collection unit is implemented by the control unit 46A of the head-mounted terminal 314 or the specific processing unit 290 of the data processing device 12. For example, the acquisition unit uses the camera 42 or communication I / F 44 of the head-mounted terminal 314 to acquire step data, which is then processed by the specific processing unit 290 of the data processing device 12. For example, the analysis unit is implemented by the specific processing unit 290 of the data processing device 12, which analyzes the data from the collection unit and the acquisition unit. For example, the generation unit is implemented by the specific processing unit 290 of the data processing device 12, which uses a generation AI to generate a menu. For example, the serving unit is implemented by the speaker 240 and display 343 of the head-mounted terminal 314 or the specific processing unit 290 of the data processing device 12, which provides the generated menu to the user. The correspondence between each unit and the device or control unit is not limited to the above examples and various changes can be made.

[0528] 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.

[0529] Fourth Implementation Method Figure 7 An example of the configuration of the data processing system 410 according to the fourth embodiment is shown.

[0530] 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.

[0531] 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).

[0532] 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.

[0533] 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.

[0534] 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 an angle equivalent to the field of vision of an average healthy person).

[0535] 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.

[0536] 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.

[0537] 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.

[0538] 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.

[0539] 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.

[0540] 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.

[0541] 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".

[0542] 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.

[0543] 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.

[0544] 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.

[0545] 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.

[0546] 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.

[0547] 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 partially or entirely performed by AI, but is not limited to this example. Furthermore, the processing performed by 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 AI including the generation AI.

[0548] 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.

[0549] 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.

[0550] 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.

[0551] 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...). 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.

[0552] 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.

[0553] 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.

[0554] 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).

[0555] 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.

[0556] 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."

[0557] 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.

[0558] 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).

[0559] 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.

[0560] 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.

[0561] 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.

[0562] 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.

[0563] 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.

[0564] 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.

[0565] 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.

[0566] 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.

[0567] 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.

[0568] 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.

[0569] In addition, the following notes are provided in response to the above explanation.

[0570] Example 1 (Note 1) An information processing system, characterized in that it comprises: An apparatus for acquiring structured data from a personnel information processing device, a business evaluation processing device, a target management processing device, a business progress analysis processing device, and a learning processing device via a communication interface, and for integrating the structured data based on an identifier and storing it in a data storage device. A device for performing a data parsing program on the structured data stored in the data storage device to perform missing value completion, standardization, summarization and feature generation, thereby generating a set of parsed data containing evaluation indicators, target achievement indicators and business progress indicators for operators; An apparatus for learning a prediction model using a prediction model building program based on the parsed data set, and using the prediction model to estimate the probability of target achievement and business efficiency indicators of operators, and storing the estimated results as estimated result data in the data storage device. A device for generating visualization data containing time change information, achievement probability information and intra-group ranking information based on the parsed data set and the inferred result data using a visualization generation program, and outputting the visualization data as visualization information that can be displayed on a display device. An apparatus for obtaining, based on user identification information, the parsed data set and the inference result data related to the user from the data storage device, summarizing and extracting the obtained data to generate background information containing the user's business status, prediction results and comparison information with others; An apparatus for automatically generating prompt statements containing analysis request content, object scope, and output format conditions based on the background information and instruction rules for generative artificial intelligence models; An apparatus for combining the prompt statement with the background information to form input information for a generative artificial intelligence model, calling the generative artificial intelligence model, performing analysis or prediction processing based on the input information, and obtaining a generation result containing suggestion information and explanatory information. A device for outputting the generated results in association with the visualization information, and for providing users with support information for developing or revising business plans based on the associated output.

[0571] (Note 2) The information processing system according to Appendix 1 is characterized in that, The device for generating background information is configured to use a group information processing program to statistically process the evaluation indicators, target achievement indicators, and business progress indicators of other members to generate baseline distribution information, and to derive relative evaluation information based on the baseline distribution information and the user's indicator values, generate a prompt statement containing the relative evaluation information, and input the prompt statement into the generative artificial intelligence model to obtain the generation results of suspected evaluations and improvement schemes based on the relative evaluation information.

[0572] (Note 3) The information processing system according to Appendix 1 is characterized in that, The system is configured to use a time-based control program to repeatedly generate the background information, automatically generate the prompts, and call the generative artificial intelligence model at predetermined cycles. The generation results obtained in each cycle are stored in chronological order, and the generation results are provided to the user terminal as notification information, thereby providing continuous incentives and growth management support based on changes in user behavior and achievement status.

[0573] Application Example 1 (Note 1) An information processing system, characterized in that it comprises: Devices for acquiring human resources information; A device for acquiring business progress information; A device for acquiring target information; An apparatus for integrating the human resources information, the business progress information, and the target information to generate a data set, and for calculating statistics and features based on the data set; Apparatus for generating prompt statements for input to a generative artificial intelligence model based on the statistics and the features; An apparatus for inputting the prompt statement and the data set into the generative artificial intelligence model to obtain generated results of work steps or suggestions related to improving work efficiency; An apparatus for dividing the generated result into multiple short texts suitable for display on a portable display terminal, formatting the multiple short texts and adding display order information to generate terminal-side display data; A device for transmitting terminal-side display data to a wearable visual display terminal via a communication network; A device for sequentially displaying the work steps or suggestions contained in the terminal-side display data on the wearable visual display terminal, and switching the display content according to the user's input operation; An apparatus for generating log information on the wearable visual display terminal based on the user's execution status and input operations for each task step, and for re-sending the log information to the server as part of the human resources information, the business progress information, and the target information to be added to the data set.

[0574] (Note 2) The information processing system according to Appendix 1 is characterized in that the system further includes: means for generating a prompt statement for calculating a relative evaluation index of business progress for a user in the generative artificial intelligence model based on the human resources information and business progress information related to other members; and means for determining the relative evaluation index according to the generation result output by the generative artificial intelligence model and including the relative evaluation index in the terminal-side display data for prompting on the wearable visual display terminal.

[0575] (Note 3) The information processing system according to Appendix 1 is characterized in that the system further includes: a device for periodically acquiring the human resources information, the business progress information and the target information over time, updating the prompt statement in each period and inputting it into the generative artificial intelligence model, thereby repeatedly providing the updated work steps, suggestions and evaluation indicators to the wearable visual display terminal to continuously motivate the user.

[0576] Example 2 (Note 1) An information processing system, characterized in that it comprises: A means for obtaining personnel-related evaluation information from an information storage device by a processing unit in an information processing device; Means for preprocessing the evaluation information by the processing unit, the preprocessing including missing value imputation, outlier removal and statistical standardization to generate standardized evaluation data; A means for calculating the relative pseudo-evaluation of each group component by having the processing unit perform neighborhood search-based classification processing based on the standardized evaluation data and evaluation data of other group components. Means for generating prompt statements by the processing unit based on the relative pseudo-evaluation and the evaluation data, and including the evaluation data and the relative pseudo-evaluation in the prompt statements; A means for the processing unit to cause the generative artificial intelligence model to perform analysis or prediction processing according to the prompt statement, and to associate the output result with the relative pseudo-evaluation to generate visual data; Means for sending the visualization data from the processing unit to a display device, and presenting it on a user terminal as visualization information including the relative pseudo-evaluation and the output results of the generative artificial intelligence model.

[0577] (Note 2) According to the information processing system described in Appendix 1, the processing unit periodically generates the prompt statement based on time information and classification information of group constituent elements, updates the relative pseudo-evaluation and the output results of the generative artificial intelligence model at various times, and provides them as the visualization information to support the improvement of user motivation and the growth management of group constituent elements.

[0578] (Note 3) According to the information processing system described in Appendix 1, the processing unit calculates the overall statistical indicators of the group from the standardized evaluation data, and calculates the relative pseudo-evaluation as at least one of the difference indicator, ranking indicator or quantile indicator for the statistical indicators, and embeds the indicator into the prompt statement to improve the accuracy of the generative artificial intelligence model in generating explanatory text or suggestion content.

[0579] Application Example 2 (Note 1) An information processing system, characterized in that it comprises: The personnel-related information processing unit is configured to acquire and integrate personnel-related attribute information, evaluation information, target information, and progress information. The business-related information processing unit is configured to acquire and integrate operational information, production information, fault information, and operational status information related to the work object; The emotion-related information processing unit is configured to acquire the user's emotional state information and parse the emotional state information to generate emotional evaluation indicators. The comprehensive data generation unit is configured to generate a comprehensive dataset using the personnel-related information, the business-related information, and the sentiment evaluation index, and to calculate the performance indicators of personnel and work objects based on the comprehensive dataset; The prompt statement generation unit is configured to generate prompt statements based on the comprehensive dataset and the performance indicators to instruct the generative artificial intelligence model to perform analysis or prediction on the evaluation, relative comparison, efficiency improvement or task priority adjustment of personnel or work objects. The generative artificial intelligence model linkage unit is configured to input the prompt statement and the comprehensive dataset into the generative artificial intelligence model, and obtain analysis results, pseudo-evaluations and suggestion information from the generative artificial intelligence model; The visualization processing unit is configured to use the performance indicators and the analysis results to generate display data for visually representing changes in personnel evaluations, relative positions with other members, efficiency ranking of work objects, the relationship between emotional state and performance, and the recommendation information. The information providing unit is configured to provide the displayed data to the output device, enabling users and managers to perform business organization, change the setting of work objects, or adjust the task priority based on the pseudo-evaluation and the suggestion information.

[0580] (Note 2) The information processing system according to Appendix 1 is characterized in that, The prompt statement generation unit is configured to generate prompt statements based on the progress information and performance indicators of multiple members or multiple task objects, to instruct the generative artificial intelligence model to calculate the relative pseudo-evaluation of each member or task object; the generative artificial intelligence model linkage unit is configured to obtain the relative pseudo-evaluation calculated based on the prompt statement; and the visualization processing unit is configured to use the relative pseudo-evaluation to generate sorting information and relative position information as display data.

[0581] (Note 3) The information processing system according to Appendix 1 is characterized in that, The prompt statement generation unit is configured to periodically generate prompt statements within a predetermined period, based on the comprehensive dataset and the sentiment evaluation index, to instruct the generative artificial intelligence model to perform pseudo-evaluation calculations that take into account time changes and to provide suggestions for improving motivation; the generative artificial intelligence model linkage unit is configured to acquire the pseudo-evaluation and suggestion information calculated based on the prompt statements; and the information providing unit is configured to continuously provide the pseudo-evaluation and suggestion information to users and managers to support motivation improvement and growth management.

Claims

1. An information processing system, characterized in that, include: processor; The processor is configured to link the personnel management device, performance evaluation device, target setting device, business progress analysis device and machine learning device to analyze employee evaluations, targets and business progress. The processor is configured to generate prompts that instruct the generative artificial intelligence model to perform specific analyses or predictions; The processor is configured to enable a generative artificial intelligence model to perform analysis or prediction based on the prompt information, and to visualize the results of the analysis or prediction.

2. The information processing system according to claim 1, characterized in that, The processor is configured to generate a prompt message for calculating a relative suspected rating based on the business progress of other department members, and to enable a generative artificial intelligence model to calculate a relative suspected rating based on the prompt message.

3. The information processing system according to claim 1, characterized in that, The processor is configured to periodically generate prompts, calculate suspected evaluations based on the prompts, and provide the suspected evaluations to the user in order to increase the user's motivation and facilitate managers in managing the growth of their subordinates.

Images