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: receive information through an interface for receiving information from a user; generate a prompt for indicating that a generative artificial intelligence model parses the received information and generates a relevant use case; and based on the generated use case, apply a pre-set rule to sort the information to a corresponding department or service.

Description

Technical Field

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

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

[0003] In existing technologies, customer service and information processing systems designed for internal enterprise or multi-departmental collaboration typically categorize and distribute information based solely on keywords or business categories of user-submitted content. This approach struggles to timely and accurately understand users' specific needs and potential demands within real-world business scenarios. Especially in complex business environments, users often express their problems and needs through free text, voice descriptions, or videos containing facial expressions. Existing systems lack the ability to automatically generate relevant use cases tailored to specific scenarios from this unstructured data using generative AI models, resulting in insufficient accuracy in information sorting and routing, and low processing efficiency. Furthermore, existing systems generally ignore the impact of user emotional states on service priority and processing order, failing to dynamically adjust information processing priorities based on factors such as user emotional stress and urgency. This leads to significant shortcomings in service experience, complaint handling, and crisis response. Therefore, a system is needed that combines generative AI use case generation capabilities with user emotion analysis capabilities to improve the intelligence of information sorting and the agility and personalization of user services. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an information processing system. The system includes a processor configured to: receive information through an interface for receiving information from a user; generate prompts instructing a generative artificial intelligence model to parse the received information and generate relevant use cases, enabling the generative artificial intelligence model to automatically construct use case descriptions matching specific business scenarios based on user-input voice, text, or other unstructured data; and, based on the generated use cases, apply pre-defined rules to sort the information to appropriate departments or services, thereby achieving intelligent routing and refined distribution of user requests. Furthermore, the processor is also configured to: analyze the user's emotional state using voice data, text data, and facial expression data as input; evaluate the user's emotions and generate an emotion analysis result; and, based on the emotion analysis result, adjust the priority of information, calculate a priority score, and determine the processing order of information according to the priority score. Through these technical means, this invention can understand user business needs while comprehensively considering the urgency of user emotions, achieving dynamic optimization of information sorting and processing priorities, thereby improving the overall service quality and response efficiency of the system.

[0005] A "system" refers to a comprehensive whole that includes at least one processor and an interface for interacting with users, and is a collection of technical solutions for performing functions such as information reception, parsing, use case generation, information sorting, and priority control.

[0006] A "processor" is a hardware or virtual processing unit that can execute program instructions and perform calculations, control, and logical processing on input data. It can be a single physical processor, a processing cluster consisting of multiple processors, or a virtual computing resource deployed in a cloud environment.

[0007] An "interface" refers to a collection of hardware and software components used for information exchange between a user and a system. These components can include graphical user interfaces, command-line interfaces, application programming interfaces (APIs), voice input interfaces, sensor input interfaces, etc., through which the user's voice, text, facial expression data, or other forms of information are received.

[0008] "Information" refers to a collection of data related to user needs, problem descriptions, feedback, and inquiries submitted by users through interfaces or collected by the system, including but not limited to text data, voice data, emoticon data, images, or combinations thereof.

[0009] "Generative artificial intelligence models" refer to artificial intelligence models built based on machine learning or deep learning techniques that can automatically generate corresponding output content based on input data. These include, but are not limited to, large language models and multimodal generative models, which are used to generate relevant use cases or text descriptions from user input information.

[0010] "Hints" refer to instructional or guiding data generated by the processor and input into the generative artificial intelligence model. They are used to limit or guide the generative artificial intelligence model to parse the received information and generate relevant use cases, including instruction text, context information, constraints, and expected output format.

[0011] "Relevant use cases" refer to application scenario descriptions or solution examples generated by generative artificial intelligence models based on user input information, which correspond to specific business scenarios or needs. They are used to help the system determine which department or service the information should be sorted to for processing.

[0012] "Predefined rules" refer to a set of logical rules defined in advance by administrators or designers during the system deployment or configuration phase. These rules are used to sort and route information based on conditions such as use case content, business type, and urgency. They may include matching rules, priority rules, business process rules, etc.

[0013] "Sorting" refers to the process by which a processor, based on generated relevant use cases and pre-defined rules, categorizes, filters, and routes received information to determine which department, service, or processing queue to send the information to.

[0014] A "department" refers to a business unit within an organization or enterprise that performs specific functions, such as customer service, technical support, sales, and legal departments, which are used to receive and process information sorted by the system.

[0015] "Service" refers to specific functional modules or business processes provided by a system or organization that are oriented towards user needs, such as online customer service, fault handling service, complaint handling service, consultation and answering service, etc.

[0016] "Voice data" refers to audio signals provided by users through microphones or other sound acquisition devices, which are then digitized and used by the system for speech recognition and emotion analysis.

[0017] “Text data” refers to text information that is input by the user or recognized by the system, including text entered through the keyboard, text converted from speech recognition results, and text messages obtained from the messaging system.

[0018] "Face data" refers to data that reflects a user's facial expression state, including facial feature information collected by a camera and processed through image processing, or face analysis results output by a third-party face recognition module.

[0019] "Emotional state" refers to the psychological and emotional state exhibited by a user at a specific moment of interaction, such as anger, anxiety, tension, calmness, satisfaction, etc., which is obtained by the system based on a comprehensive evaluation of voice data, text data, and facial expression data.

[0020] "Emotion analysis results" refer to the quantitative or classification results of the user's emotional state obtained by the processor after analyzing the input voice data, text data, and facial expression data, including parameters such as emotion category, emotion intensity, and sense of urgency.

[0021] "Priority" refers to the importance of a piece of information relative to other information in the system's processing order, and can be dynamically adjusted based on business rules and sentiment analysis results.

[0022] "Priority score" refers to a numerical or hierarchical index calculated by the processor based on factors such as sentiment analysis results, business type, and user level. It is used to quantify the processing priority of information and serve as the basis for determining the processing order.

[0023] "Processing order" refers to the specific execution order determined by the system after sorting multiple pieces of information to be processed. It is used to arrange the processing position of each piece of information in the queue in order to achieve reasonable resource allocation and response time optimization. Attached Figure Description

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

[0058] In existing computer information processing technologies, the automatic collection, parsing, and utilization of publicly available information related to wireless communication technology faces the following technical challenges. First, conventional data crawling and text analysis systems typically only perform keyword matching or simple classification, failing to create an efficient organizational index structure within the computer. Consequently, they struggle to accurately extract truly "active" organizations in wireless communication technology applications and their corresponding overseas application examples from massive amounts of heterogeneous publicly available information, resulting in insufficient accuracy in downstream analysis and decision support. Second, traditional human-computer interaction methods rely heavily on operators manually writing query statements or natural language questions. The computer system itself lacks the ability to dynamically generate high-quality prompts based on the latest analysis results. This leads to semantic non-focus and insufficient constraints in generative artificial intelligence models when receiving input, thereby reducing the relevance and usability of the model's generated results. Furthermore, existing systems, when utilizing response information output by generative artificial intelligence models, often present it to users as a one-time result. They fail to unify and persistently manage the response information with previous evaluation and organizational index information at the system level. The computer cannot accumulate existing responses into reusable knowledge resources, resulting in each interaction being in a "cold start" state, making it difficult to optimize overall reasoning efficiency and response quality over time. Moreover, when the number of wireless communication application examples across different industries and technologies increases dramatically, the system lacks a mechanism to automatically adjust query priorities based on internal computer evaluation indicators. It cannot adaptively highlight high-value technologies or industry directions when generating prompts, leading to a waste of computing and communication resources. In summary, there is an urgent need for a solution that improves computer technology from both system architecture and data processing flow perspectives. This solution should introduce structured organizational index construction, generative artificial intelligence model-driven semantic evaluation, and a dynamic prompt generation mechanism based on indicator values ​​between the server and communication devices. This would enable the computer to automatically and continuously optimize the retrieval, modeling, and interaction processes of wireless communication technology application examples, thereby significantly improving the automation, accuracy, and computing resource utilization efficiency of information processing.

[0059] 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.

[0060] In this invention, the server includes: means for receiving identification information representing the location of publicly available information about an object from a communication device; means for communicating with an external information processing device based on the identification information and automatically acquiring standardized descriptive information from multiple information sources; means for parsing the structure of the descriptive information and extracting textual information from unstructured information to generate a sequence of information elements; and means for performing lexicalization and encoding preprocessing on the sequence of information elements and converting it into feature information for input to a generative artificial intelligence model; the server also includes: means for performing generative artificial intelligence model inference processing using the feature information to generate evaluation information representing the relevance of wireless communication technology to specific wireless communication technology and organizational attributes; and means for performing evaluation information based on the evaluation information to generate evaluation information related to specific wireless communication technology and organizational attributes. The server includes: an apparatus for actively utilizing an organization and generating an organization index information representing the correspondence between the organization and overseas examples; an apparatus for selecting a sentence structure pattern for automatically generating prompts for input to a generative artificial intelligence model based on the organization index information and the evaluation information; an apparatus for generating query text containing at least one of a technology type, industry type, and region type; an apparatus for sending the query text to a communication device and generating control information for inputting the query text to the generative artificial intelligence model based on input operations from the communication device; and an apparatus for storing and accumulating response information obtained from the generative artificial intelligence model in correspondence with the organization index information and the evaluation information as knowledge information that can be reused in subsequent query text generation. This allows for the formation of a closed-loop mechanism within the computer system for information acquisition, semantic analysis, index construction, and interaction optimization. The server can automatically identify and structurally represent high-value application examples and active organizations in wireless communication technology. Combined with generative artificial intelligence models, it can perform deep semantic evaluation of public information, dynamically generate high-quality prompts for the model, and precipitate the results of multiple rounds of interaction into reusable knowledge resources. This significantly improves the relevance and accuracy of information retrieval and question-answering processes, enhances the input quality and output performance of generative artificial intelligence models, improves the overall utilization efficiency of computing and communication resources, and achieves substantial improvements in computer technology.

[0061] "Communication device" refers to a terminal device that can send and receive data with a server through wired or wireless communication networks, including but not limited to computer terminals, mobile terminals, tablet terminals, and other information processing devices with network communication functions.

[0062] "Identification information" refers to information used to uniquely or quasi-uniquely indicate the location or source of publicly available information, including but not limited to Uniform Resource Locator (URL), Uniform Resource Identifier (URLI), index number, or other marker data used to retrieve target information from external information processing devices.

[0063] "Public information" refers to information that can be accessed through public networks or open interfaces without individual authentication, including but not limited to web page content, publicly available technical documents, news reports, and publicly disclosed business introduction data.

[0064] "External information processing device" refers to an information processing system or server located outside this system that can provide information services or data access interfaces, including but not limited to network service systems, data provision platforms and search service systems.

[0065] "Information provider" refers to any entity in the set of data sources that store or provide publicly available information, including but not limited to websites, database services, or web application services.

[0066] "Standardized descriptive information" refers to descriptive information obtained from multiple information sources and processed with unified format and standardized fields, used to achieve a unified data structure representation across different sources.

[0067] "Description structure" refers to the markup structure, hierarchical structure, or formatting structure contained in the description information, including but not limited to markup language structure, field layout structure, and document hierarchy structure.

[0068] "Unstructured information" refers to information that does not conform to a predefined data architecture and is difficult to access directly through field indexes, including but not limited to natural language text, free-form documents, and free text portions in mixed content pages.

[0069] “Textual information” refers to a sequence of characters that can be processed by a computer, extracted from descriptive or unstructured information, including but not limited to natural language sentences, word sequences, or symbol sequences.

[0070] "Information element sequence" refers to an ordered set composed of multiple extracted text information units arranged in a predetermined order, with each text information unit representing a basic information fragment used for subsequent semantic analysis.

[0071] "Generative artificial intelligence models" refer to machine learning models that can automatically generate text, vectors or other data outputs based on input feature information, including but not limited to deep learning-based language models, sequence generation models and their variants.

[0072] "Lexical processing" refers to the preprocessing operation of decomposing continuous textual information into words, subwords, or tokenized units before inputting the sequence of information elements into a generative artificial intelligence model. This includes, but is not limited to, word segmentation, subword splitting, and tokenization.

[0073] "Encoding processing" refers to the process of converting lexicalized tokenized units into numerical representations that can be received by the model, including but not limited to index mapping, vector embedding, and positional encoding.

[0074] "Feature information" refers to a set of numerical data that represents the semantic or structural characteristics of the input content, obtained by lexicalization and encoding of a sequence of information elements, and is used as input for generative artificial intelligence models.

[0075] "Inference processing" refers to the forward computation process performed in a generative artificial intelligence model based on input feature information to obtain output results or intermediate representations, including but not limited to operations such as classification inference, relevance calculation, and text generation.

[0076] "Evaluation information" refers to the result data generated by the inference processing of generative artificial intelligence models, which is used to quantitatively or qualitatively represent the relevance between input information and the use of wireless communication technology in cases and organizational attributes, including but not limited to relevance scores, tag sets, or ranking indicators.

[0077] "Organizational attributes" refer to characteristic information related to an organizational entity, including but not limited to industry category, geographical information, technology usage, and business area.

[0078] "Organizational index information" refers to structured index data built based on evaluation information, used to represent the correspondence between organizations and corresponding utilization cases, and used to support the retrieval and associated access to organizations and their related cases.

[0079] "Prompt statements" refer to natural language expressions that are automatically generated for input into generative artificial intelligence models. They are used to guide or constrain the model's generative behavior and include, but are not limited to, text content used for asking questions, giving instructions, or setting context.

[0080] A “sentence structure pattern” refers to a predefined language structure template used to construct prompt statements, which specifies the arrangement and grammatical combination of information elements in the prompt statement.

[0081] "Technology category" refers to the classification information of technologies involved in system processing, including but not limited to wireless communication standard categories, network technology categories, and related technical field categories.

[0082] "Industry category" refers to industry category information obtained by classifying organizations according to economic activities or business areas, including but not limited to manufacturing, service, transportation and other industry sectors.

[0083] "Region type" refers to the classification information that represents the geographical area involved in the organization or use case, including but not limited to national-level regions, regional alliance-level regions, and city-level regions.

[0084] "Question text" refers to natural language text generated based on sentence structure patterns and organizational index and evaluation information, used to pose questions or instructions to generative artificial intelligence models.

[0085] "Control information" refers to the instructional data used to control the input process of query text into the generative artificial intelligence model, including but not limited to calling parameters, session identifiers, and routing information.

[0086] "Response information" refers to the response output of a generative artificial intelligence model after receiving a query text, including but not limited to natural language answers, structured conclusions, and suggestions.

[0087] "Knowledge information" refers to a data set that is accumulated by storing response information in correspondence with organizational index information and evaluation information, and can be reused in subsequent query text generation. It is used to support knowledge reuse and reasoning optimization within the system.

[0088] In this embodiment of the invention, the server undertakes the core data processing and model inference functions, while the terminal undertakes the human-computer interaction and prompt display functions, allowing users to perform selection and confirmation operations through the terminal. The system of this invention can be deployed on a cluster consisting of multiple physical servers, each equipped with a central processing unit (e.g., a multi-core general-purpose processor), a graphics processing unit (e.g., a general-purpose parallel computing accelerator), main memory, and a non-volatile storage device based on solid-state memory. The server runs application frameworks and deep learning frameworks (e.g., open-source frameworks for tensor computation and neural network inference) on an operating system (e.g., a Unix-like operating system), and executes data fetching and preprocessing programs through a scripting language runtime environment. The terminal can be a smartphone, tablet, or desktop terminal, running a terminal operating system (e.g., a mobile operating system or a desktop operating system), and communicates with the server via network protocols.

[0089] In one implementation, the server uses a web crawling framework (such as Scrapy in Python) and a network request library to send requests via Hypertext Transfer Protocol (HTTP) or Secure Hypertext Transfer Protocol (SSL) over a physical network interface to obtain public information. The server stores the obtained Hypertext Markup Language (HTML) pages in non-volatile storage and loads the pages into main memory for parsing. The server uses a document parsing library (such as an HTML or XML parsing library) to traverse the markup structure, extracting text from title tags, body container tags, and structured tags. The server uses regular expression matching and named entity recognition models to extract candidate strings such as organization names, geographical names, and technical names from the text, and organizes the extracted text information into a sequence of information elements by field. The server stores the sequence of information elements in key-value pairs in a relational database or document database, where each record includes at least Uniform Resource Locator (URI) information, title text, body text, candidate organization name, candidate region information, and an acquisition timestamp.

[0090] When performing semantic analysis on the information element sequence, the server uses a generative artificial intelligence model. In a preferred embodiment, the server employs a language model with an encoder-decoder structure based on a self-attention mechanism or a decoder-only structure as the generative artificial intelligence model. The server sets up multiple layers of self-attention sublayers, feedforward network sublayers, and normalization sublayers in the model structure, with features transferred between layers via residual connections. On the input side, the server uses a lexicalization module to divide the text information into sub-word tags, maps each sub-word tag to an integer identifier using a vocabulary, and then converts the integer identifiers into dense vectors through an embedding matrix, while simultaneously superimposing positional encoding vectors to form feature information. The server inputs the feature information into the model's multi-layer self-attention structure, calculates the relevance score between features at each position through a scaled dot product attention mechanism, and obtains a weighted contextual feature representation. At the output layer, the server uses a fully connected layer and an activation function to map high-dimensional features into classification scores or relevance scores, used to represent the relevance between a descriptive piece of information and examples of wireless communication technology utilization and organizational attributes.

[0091] In training the generative AI model, the server utilizes supervised learning methods in one implementation. The server prepares a labeled training dataset where each sample includes a text fragment and labeled technical category, industry category, binary label indicating whether it is a wireless communication application instance, and an organization activity score. During training, the server optimizes the classification output using a cross-entropy loss function and performs regression optimization using a mean squared error loss function. The server combines these losses in a weighted manner into a total loss function and iteratively updates the model parameters using stochastic gradient descent or its variants (e.g., momentum optimization algorithms or adaptive learning rate optimization algorithms). The server may employ data augmentation techniques during training, such as synonym replacement, sentence order perturbation, or masked language modeling, to improve the model's robustness to diverse texts. The server propagates error signals between layers of the neural network using a gradient backpropagation algorithm, updating the self-attention weight matrix, feedforward network weight matrix, and embedding matrix parameters, thereby enabling the model to accurately output evaluation information after training.

[0092] During the inference phase, the server uses a trained generative artificial intelligence model to process the newly collected information element sequence. The server receives feature information output from the preprocessing module and calls the inference interface of the deep learning framework to perform forward computation on the graphics processing unit (GPU) to improve inference speed and reduce the burden on the central processing unit (CPU). The server extracts tags related to wireless communication technology utilization examples from the model's classification results, reads organization activity scores from the output regression results, and analyzes the correlation between text fragments and technical keywords from the output attention weight distribution. The server organizes these outputs into evaluation information and generates one or more technology category tags, industry category tags, and region category tags for each record. The server then constructs organization index information based on the evaluation information. The server aggregates records with the same organization name or the same standardized organization identifier, calculates the average or weighted average of their activity scores, and forms an organization activity index. The server constructs index entries for organization name, activity index, main technology categories, main industry categories, and corresponding example link lists, and creates inverted indexes for fields such as organization name, technology category, and industry category in search engines or databases. Through this specific index structure, the server can quickly locate the target organization and its related events in subsequent queries using hash lookups or inverted index searches, thereby improving retrieval speed and reducing access latency.

[0093] When generating prompts, the server selects a sentence structure pattern based on organizational index information and evaluation information. In one implementation, the server predefines multiple sentence structure patterns, such as patterns for case collection, comparative analysis, and solution conception. The server selects a suitable pattern template based on a combination of the current target technology type, industry type, and region type, and fills the template placeholders with the technology name, industry name, and region name extracted from the organizational index information to generate natural language query text. In another implementation, the server can encode the organizational index information into structured prompts, inputting them along with fixed instruction text into the decoding part of the generative AI model, allowing the model to generate more flexible natural language prompts. When generating prompts, the server can dynamically adjust the frequency and order of different technology types or industry types based on evaluation information. For example, when the activity index of a certain technology type is significantly higher than other technology types, the server moves that technology type to the front of the sentence structure pattern or adds reinforcement descriptions to guide the generative AI model to focus on that technology area. This evaluation metric-driven template selection and filling strategy enables prompt statements to adaptively reflect the latest analysis results, improving the focus and effectiveness of model input.

[0094] When the server collaborates with the terminal, it sends the generated query text to the terminal via a network interface. The terminal receives and caches the query text in its local storage. The terminal displays multiple candidate prompts in a list on its display device for the user to select or edit. For example, the terminal may display the following prompts: "Based on typical cases of 5G use in overseas manufacturing over the past three years, please analyze the differences in network architecture and application scenarios among different companies in the construction of smart factories." "Please list representative companies overseas that use LPWA (such as NB-IoT, LoRaWAN) in logistics tracking, and compare their differences in coverage, power consumption and cost." "Please compare typical cases of 5G smart logistics applications in Germany, Japan, and the United States over the past five years, and analyze the differences in policy support and business models among different countries." After receiving touch or pointer input from the user, the terminal performs string editing on the selected prompt statement, inserting supplementary conditions entered by the user into predetermined positions or replacing parts of the phrase, such as replacing "smart factory" with "smart city," or adding a time constraint of "the last five years." After determining the final prompt statement, the terminal sends it back to the server as text data via a network protocol, along with the user's identifier and session identifier. Upon receiving the prompt statement, the server encapsulates it as an input request for a generative artificial intelligence model, specifying inference hyperparameters such as maximum generation length and temperature parameters, and then calls the model service to generate the response. After obtaining the response information from the model service, the server establishes a mapping relationship between the response information and the corresponding organizational index information and evaluation information, and writes it into the knowledge base. In one implementation, the server uses a graph structure to store knowledge information, connecting organizational nodes, instance nodes, and response summary nodes through edges, and assigning relevance weights to each edge. In this way, when generating prompt statements in the future, the server can traverse nodes in the knowledge graph that are highly relevant to the target technology or industry type, prioritizing the use of information from these nodes to construct new prompt statements, thereby achieving knowledge reuse and shortening the inference path.

[0095] Through the specific data structures and algorithmic processes described above, the server not only simply achieves information retrieval and question answering, but also performs multi-level structuring and indexing of information internally. Combined with a generative artificial intelligence model, it implements an evaluation-driven prompt generation mechanism. This mechanism differs from traditional methods based on fixed query templates or entirely manual query writing by users. Internally, the server uses evaluation information generated by the model as weights to sort and select different technology and industry categories, comprehensively considering activity index, case quantity, and time distribution in a non-linear function manner, thereby dynamically adjusting the content and focus of the prompts. This specific set of rules and processing order is not a simple automation of conventional manual retrieval steps, but rather utilizes high-dimensional features and relevance scores output by neural networks to construct new control signals, enabling the computer to adaptively optimize subsequent inputs. This adaptive optimization brings several technical benefits: firstly, at the model inference stage, the server reduces unnecessary generation space by decreasing the participation of irrelevant or low-weight technology labels, improving the relevance of generated results and reducing the time users need to spend filtering useless information; secondly, at the index building stage, the server achieves efficient retrieval by organizing the activity index, maintaining low query latency even on large-scale datasets, thereby improving overall processing speed. Furthermore, by persistently storing response information, evaluation information, and organizational index information together, the system of the present invention can reuse historical reasoning results in subsequent sessions, avoid repeated calculations, reduce the load on the graphics processing unit, and thus achieve the effects of saving computing resources and shortening response time.

[0096] In another implementation, the server can employ different generative AI model structures, such as a combination of a bidirectional encoder-based language understanding model and an autoregressive decoder-based text generation model. The server can use the encoder to perform deep semantic encoding on the sequence of information elements, extracting sentence-level and document-level features, and then use the decoder to generate prompts based on the extracted features and organizational index information. In this structure, the server can fine-tune the encoder and decoder separately, with the encoder using contrastive learning or masked language modeling objectives and the decoder using conditional language modeling objectives to further improve semantic representation and generation quality. The server can also employ a modular pipeline, combining a rule engine with neural network outputs: the server first uses the neural model to output evaluation information, then the rule engine determines whether to introduce a specific technology or industry category into the candidate set of prompts based on predefined weights and thresholds, and finally, a lightweight language generation module linearly combines the selected set into natural language text. This hybrid processing approach provides a trade-off between computational complexity and interpretability.

[0097] The terminal can employ various interface formats across different implementations. It can provide multiple areas on its graphical user interface: one area displays a list of automatically generated prompts, another displays responses from the generative AI model, and filtering controls allow users to select prompts by technology or industry. The terminal can also maintain preference parameters in local storage based on user history and include these parameters when requesting prompts from the server. The server then reorders sentence structures during prompt generation to better suit specific user groups. When receiving responses from the server, the terminal can summarize long texts, presenting only key paragraphs and providing an expand button to view the full content.

[0098] When using the system of this invention, users do not need to understand the internal computational details of the deep learning model. They only need to select or slightly edit the prompts through the terminal interface to obtain high-quality analysis results for wireless communication technology application examples. However, the improvement of computer technology in this invention is not limited to the user experience level, but is mainly reflected in the information processing method inside the server. By introducing a linkage mechanism between evaluation information and organizational index information based on generative artificial intelligence models, the server changes the traditional system's "static index + fixed query template" mode, creating a closed-loop feedback between index construction and query construction. This closed loop enables the system to continuously focus on high-value technologies and industry directions during long-term operation, gradually optimizing the knowledge base structure, and improving retrieval efficiency and reasoning accuracy over time. This is reflected in the evolution of the computer's internal data structure and algorithm behavior, rather than simply the electronification of human workflow.

[0099] Through the above embodiments, the division of labor among the server, terminal, and user is clear: the server is responsible for complex data processing, model training and inference, index construction, and prompt generation; the terminal is responsible for interaction with the user, displaying prompts and responses, and managing local configuration information; and the user guides the system to output results that better meet their needs through simple selection and editing operations. This invention is not limited to a specific hardware platform, operating system, or specific implementation language. Without departing from the technical concept defined in the claims, various software and hardware substitutions and combinations can constitute other embodiments of this invention.

[0100] use Figure 11 The processing procedure is explained.

[0101] Step 1: The server receives identification information and constructs a device task. The server takes identification information from the terminal as input, which includes one or more publicly available location identifiers (e.g., multiple network addresses). The server creates task objects in main memory and encapsulates each network address as a unit to be crawled. The server generates a crawling queue based on the identification information, enqueues each unit to be crawled, and records the task batch number and timestamp. The server performs validity checks and deduplication on the input identification information, removing malformed or duplicate network addresses, resulting in a standardized list of crawling targets as output for subsequent data acquisition.

[0102] Step 2: The server crawls publicly available information and generates raw page data. The server takes the target list output from step 1 as input and initiates network requests using the network interface at a predetermined concurrency level to retrieve the corresponding public information pages. The server communicates with the external information processing device using Hypertext Transfer Protocol (HTTP) or Secure Hypertext Transfer Protocol (SSL), reading the page content from the response message as a Hypertext Markup Language (HML) string. The server judges the response status code, retains only pages with successful responses, and generates a unique identifier for each page. The server writes the page content, target address, task batch number, and retrieval time together into non-volatile storage, forming raw page data records. The server's output is a dataset containing multiple raw page data records, which serves as input for subsequent parsing and processing.

[0103] Step 3: The server parses the original page and extracts text information. The server takes the raw page dataset output from step 2 as input, loads individual page contents sequentially into main memory, and constructs a document object model using a document parsing library. Based on preset structured rules, the server extracts text content from title tags, body container tags, and specific style tags, and uses regular expressions to remove script code, style definitions, and navigation bar text. The server further calls the named entity recognition module to scan the extracted text, identifying potential organization names, geographical region names, and technical terms. The server combines the extracted title text, body text, candidate organization name list, candidate region list, and original page identifier into an information element sequence. The server's output is a structured set of information element sequences, which serves as input for natural language preprocessing and feature generation.

[0104] Step 4: The server preprocesses the information element sequence and generates feature information. The server takes the set of information element sequences output from step 3 as input. First, it cleans the title and body text, including standardizing character encoding, removing redundant whitespace, and standardizing punctuation. The server then sends the cleaned text to the lexicalization module, which segments the text into sub-word tag sequences and maps each sub-word tag to an integer identifier sequence using a vocabulary. Next, the server encodes the integer identifier sequences and inputs them into the embedding mapping module to obtain corresponding vector representations. Simultaneously, it adds positional encoding vectors based on the sequence positions, forming an embedding matrix containing both content and positional information. The server packages the embedding matrix corresponding to each information element sequence, along with the candidate organization name and candidate region information, into a feature information object. The server's output is a set of feature information objects, which serves as input for the generative artificial intelligence model's inference.

[0105] Step 5: The server invokes a generative artificial intelligence model to perform semantic reasoning and generate evaluation information. The server takes the set of feature information objects output from step 4 as input and loads a pre-trained generative artificial intelligence model onto the graphics processing unit using a deep learning framework. The server inputs the embedding matrix into a multi-layer self-attention network, calculates the attention weights between each location, and generates high-dimensional contextual features through linear transformations and non-linear activation functions. At the output layer, the server performs multi-label classification on each piece of information using a classification head, obtaining the classification result of whether it is a wireless communication utilization example and the corresponding technology type and industry type labels; the server outputs an organizational activity score or relevance score through a regression head. The server associates these classification labels and numerical scores with the original information element identifiers to form evaluation information objects. The server's output is a set of evaluation information objects, including the relevance data between each record and wireless communication technology utilization examples and organizational attributes; this set serves as input for constructing the organizational index.

[0106] Step 6: The server builds organizational index information and calculates activity metrics. The server takes the evaluation information object set output from step 5 as input and aggregates records with the same organization name. Based on the aggregation results, the server calculates the activity index for each organization, including the average activity score, the number of technology types involved, the number of industry types involved, and the number of cases. The server combines the activity score and the number of cases into an organization activity index according to a predetermined weighting formula. The server combines the organization name, organization activity index, main technology types, main industry types, and associated case identifiers into organization index entries and writes all entries into the index storage structure. The server builds an inverted index on this storage structure for organization name, technology type, and industry type to support efficient queries. The server's output is a set of organization index information, which serves as input to the prompt statement generation module.

[0107] Step 7: The server generates prompt statements based on organizational index information and evaluation information. The server takes the organization index information output from step 6 and the evaluation information output from step 5 as input, and selects a set of priority organizations and cases based on the current target technology type, industry type, and region type. The server searches a predefined sentence structure pattern library and selects a sentence structure pattern that matches the selection results, such as a pattern for case lists, comparative analysis, or solution design. The server fills the placeholders in the sentence structure pattern with the technology name, industry name, and region name obtained from the organization index information, or concatenates them with fixed instruction text to generate prompts in natural language. When multiple candidate technology types or industry types exist, the server sorts them according to their activity index, prioritizing the most active categories to achieve dynamic adjustment. The server outputs multiple prompt texts and their corresponding control parameters, which are sent to the terminal for display and to drive the generative artificial intelligence model.

[0108] Step 8: The terminal receives and displays prompts for the user to choose from. The terminal takes the set of prompt texts output in step 7 as input, receives data from the server via the network communication interface, and caches it locally. The terminal displays multiple prompt texts in a list or grouped manner on the user interface of the display device, allowing the user to scroll through them. The terminal can collapse longer prompt texts, providing an expand button to view detailed content. The terminal configures selection buttons and editing entries for each prompt text in the interface controls, enabling users to perform selection or modification operations. The terminal's output is either the target prompt text selected by the user or the edited prompt text, which serves as input for subsequent model query requests.

[0109] Step 9: The user selects or edits the prompt and initiates a query request. Users take a set of prompts displayed on the terminal as input and select one through touch, mouse, or keyboard operations, or modify the prompt in the editing area. Users can insert specific time ranges, regions, or technical limitations into the prompts; for example, changing "typical cases of 5G use in overseas manufacturing" to "typical cases of 5G use in the Asia-Pacific manufacturing industry in the past five years." After confirming the prompt content, the user operates the send control on the terminal. The user's operation outputs a confirmed prompt text, which is uploaded to the server via the terminal as input for the question-answering stage of the generative artificial intelligence model.

[0110] Step 10: The server executes a generative artificial intelligence model to answer questions and returns response information. The server takes the confirmation prompt text output from step 9 as input, constructs a generative AI model call request, and sets inference parameters such as maximum generation length, temperature, and repetition penalty. After lexicalization and encoding, the server inputs the prompt text into the decoding structure of the generative AI model, generating response tags sequentially step-by-step until a termination condition is met. The server decodes the generated tag sequence into natural language response text, and can perform length pruning, sensitive word filtering, or formatting as needed. The server associates the response text with corresponding organizational index information and evaluation information, recording this interaction relationship in the knowledge base. The server's output is a structured response information object, containing the response text, associated organizational information, and technology category tags, which is sent to the terminal via the network.

[0111] Step 11: The terminal displays the response information and supports subsequent user operations. The terminal takes the response information object output in step 10 as input and parses out the response text and additional organization and technology tags. The terminal presents the response text on the display device in segmented, list, or table format, highlighting important technology and organization names. The terminal can provide scrolling, copying, or exporting functions, allowing users to save or forward the response content. Based on the tags in the response information, the terminal recommends new prompts generated based on the current response, such as follow-up questions targeting more specific technical fields. The terminal's output is a user-visible response display interface and possible subsequent prompt candidate statements, which can then be used as input for step 8 in a loop.

[0112] Step 12: The server accumulates knowledge and information and optimizes the generation of subsequent prompts. The server takes the set of response information objects output in step 10 as input, extracts key sentences, technical points, and organizational relationships, and establishes graph or relational connections with existing organizational index information and evaluation information. Based on newly added knowledge nodes, the server updates the organizational activity index and the association weights of technical categories, incrementally maintaining the index structure. The server analyzes user usage frequency and response quality indicators, adjusting the weights and selection strategies of sentence structure patterns to make subsequent prompts more consistent with actual usage. The server outputs an updated set of knowledge information and optimized organizational index information, which will serve as a new foundation for subsequent steps 6 and 7, thus forming an adaptive closed-loop processing flow.

[0113] 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".

[0114] As generative AI models are increasingly used in business scenarios, existing technologies generally suffer from the following problems: First, information processing systems simply call generative AI models based on user-input keywords, lacking automatic collection and structured preprocessing of publicly available internet information. This results in unstable input data quality, poor reliability and usability of generated results, and an inability to continuously output high-quality application examples for specific business scenarios (such as logistics and manufacturing). Second, existing systems typically present model output results in batch processing on a single terminal, lacking real-time interactive capabilities with wearable devices. In particular, they lack mechanisms for on-demand push and dynamic updates of application example information for frontline workers, making it difficult to promptly transform the knowledge from generative AI models into actionable operational guidelines. Third, existing systems utilizing generative AI models are mostly limited to direct text generation or question-and-answer sessions, failing to integrate external organization performance indicators and ranking results with the application example generation process. This prevents the automatic construction of a high-value application example list associated with "high-performing organizations," resulting in a lack of objective business data support for recommended content. Fourth, existing systems generally lack a closed-loop control mechanism of "user voice input—automatic generation of search strings / prompt statements—model loop calling—result push," failing to dynamically adjust prompt statements and data processing logic based on continuous user needs and feedback during operations. This results in low interaction efficiency and difficulty in accurately capturing user intent. Fifth, existing information sorting systems often use static rules or simple keyword matching when routing user information to different business departments or service modules. They fail to fully utilize application case information generated by generative artificial intelligence models to intelligently assist sorting rules, leading to insufficient accuracy and adaptability in information allocation, making it difficult to adapt to complex and ever-changing business scenarios.

[0115] Therefore, from the perspective of "improvement in computer technology," there is an urgent need for a new information processing system and its control method: This system can automatically collect and clean publicly available information on the server side, and perform multi-stage data processing by combining generative artificial intelligence models and prompting mechanisms; it can calculate business-related performance indicators and rank organizations, thereby generating application examples associated with high-performing organizations; it can link with portable or wearable terminals to push and display application examples in real time in frontline scenarios; and it can form a control loop that cyclically calls generative artificial intelligence models through user voice input, while using the generated application example information to improve information sorting and routing logic, thereby improving the overall efficiency and quality of the computer system in the information acquisition, analysis, recommendation, and distribution stages.

[0116] 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.

[0117] In this invention, the server includes a unit for acquiring and receiving user information via an input / output device; a unit for generating a retrieval string for obtaining public information from an external information source via a communication medium based on the user information and pre-set conditions; a unit for obtaining public information from the external information source based on the retrieval string and extracting text information from the public information according to predetermined information extraction conditions; a unit for generating a prompt statement for causing a generative artificial intelligence model to perform parsing processing or instance generation processing based on the user information, the extracted text information, and pre-set business conditions; a unit for inputting the prompt statement and the extracted text information into the generative artificial intelligence model and obtaining instance information and organizational information generated by the generative artificial intelligence model; and a unit for calculating performance indicators based on the organizational information and ranking organizations based on the performance indicators to generate... The system includes: a unit for providing organization list information; a unit for generating application case information applicable to user services based on the case information, the organization list information, and pre-defined business form information; a unit for sending the application case information to a terminal device including a portable display device or a wearable display device and converting the application case information into display data that can be displayed in real time on the terminal device; a unit for obtaining user voice input through the terminal device and generating additional search strings or additional prompt statements based on the voice input to perform cyclic control on the parsing processing or case generation processing of the generative artificial intelligence model; and a unit for generating a determination result based on the application case information to classify the information obtained from the user into different business categories or service categories according to a predetermined classification rule and allocating the user information according to the business category or service category based on the determination result. This allows for the construction of a comprehensive computer processing workflow on the server side, encompassing automatic collection and cleaning of publicly available external information, multi-stage invocation of generative artificial intelligence models based on prompts, organization and sorting based on performance indicators and generation of application examples, real-time information presentation in conjunction with wearable terminals, cyclical interactive control based on voice input, and intelligent information sorting based on application example information. This significantly improves the automation and processing accuracy of computer systems in information acquisition, analysis, recommendation, and routing, enhances information utilization efficiency in frontline business scenarios, and achieves improvements in computer technology itself.

[0118] "Information processing device" refers to an electronic device with computing and storage functions, used to execute programs and process input data, including servers, computer terminals, or information processing systems consisting of multiple processing nodes.

[0119] The "processing unit" refers to the computing unit configured in an information processing device, which is used to execute program code, perform calculations, control, and logical judgments on input data. It can be implemented by a central processing unit, graphics processing unit, or other computing circuits.

[0120] "Input / output devices" refer to hardware components used for data interaction between users and information processing devices, including display devices, keyboards, touch panels, microphones, cameras, speakers, and communication interfaces.

[0121] "User information" refers to data related to business or query needs provided by users through input / output devices and received by information processing devices, including text information, voice information, image information, and metadata.

[0122] "Communication medium" refers to the wired or wireless communication environment used to transmit data between an information processing device and an external information source, terminal device, or generative artificial intelligence model, including wired networks, wireless local area networks, cellular mobile communication networks, and the Internet.

[0123] "External information sources" refer to publicly or semi-public data providers located outside of information processing devices and accessible through communication media, including websites, database services, information aggregation platforms, and other network resources.

[0124] "Public information" refers to text, images, tables, or structured data that can be obtained from external information sources and is publicly released on the internet or other media, and is generally accessible without being limited to specific users.

[0125] A "search string" is a sequence of characters generated by an information processing device according to predetermined rules for performing search processing in external information sources. It includes keywords, logical operators, and constraints for limiting time, region, or domain.

[0126] "Information extraction criteria" refers to the set of rules or parameters used when extracting text information from public information, which are used to determine the range of text, field types, or content characteristics that need to be retained or filtered.

[0127] “Textual information” refers to natural language text data extracted from publicly available information and used for subsequent processing, including titles, body paragraphs, explanatory text, and annotation information.

[0128] "Business conditions" refer to the business-related restrictions or parameters referenced when generating prompts or using case information based on user information and text information, including target industry, business process type, scale conditions, and performance requirements.

[0129] "Business model information" refers to comprehensive information that represents the structure and operational characteristics of a user's business, including industry category, organizational size, business model, work process, and the technical architecture adopted.

[0130] "Generative artificial intelligence models" refer to statistical or neural network models built on machine learning and deep learning technologies that can generate text, code, or other content based on input data, including pre-trained language models and fine-tuned dedicated generative models.

[0131] "Prompt statements" refer to the input text constructed by the information processing device and sent to the generative artificial intelligence model. They are instructional descriptions that instruct the model to perform specific parsing or event generation processes, including task descriptions, output format requirements, and constraints.

[0132] "Analysis processing" refers to the process by which generative artificial intelligence models understand, classify, summarize, extract, or reason about input textual or user information in order to obtain structured information or semantic results.

[0133] "Case generation processing" refers to the process by which generative artificial intelligence models generate application cases, case descriptions, or solution suggestions related to specific application scenarios based on input text information, user information, and business conditions.

[0134] "Case information" refers to structured or semi-structured information obtained by generative artificial intelligence models through parsing or case generation processes, which describes the application of technology or the way business is carried out, including case background, technology used, implementation steps and effect evaluation, etc.

[0135] "Organizational information" refers to attribute data associated with event information, used to represent the entity organization participating in or implementing the event, including organization name, industry category, location, and related business indicators.

[0136] "Performance metrics" refer to numerical or metric data used to quantify an organization's business performance or economic condition, including sales, profitability, cost metrics, efficiency metrics, and other statistics that can be used for ranking and comparison.

[0137] "Organization list information" refers to the collection of information on multiple organizations arranged in a certain order, generated based on performance indicators and calculated and sorted. This includes organization identifiers, performance indicator values, and rankings.

[0138] "Utilizing case information" refers to the application solutions or operational guidelines generated based on case information and organizational list information, combined with business model information, that can be applied to specific user business scenarios. These include objectives, technologies used, implementation steps, and expected results.

[0139] "Terminal device" refers to a client device that is connected to an information processing device via a communication medium and is used to present information to a user and receive user input, including portable terminals, wearable terminals, or fixed terminals.

[0140] "Portable display devices" refer to terminal devices that users can carry with them and use to display data, including smartphones, tablets, and portable display terminals.

[0141] "Wearable display devices" refer to devices that users can wear on their bodies to present display data in their field of vision, including smart glasses, head-mounted display devices, and other wearable display terminals.

[0142] "Display data" refers to structured or semi-structured data that has been processed by information processing devices or terminal devices and is suitable for presentation on display devices in the form of graphics, text, or icons.

[0143] "Voice input" refers to the audio signals provided by a user to a terminal device or information processing device through a microphone or other audio acquisition device, used to express query intentions, control commands, or feedback information.

[0144] "Loop control" refers to the control process that automatically or semi-automatically regenerates the search string or prompt statement when new user input or new external data is obtained, and repeatedly calls the generative artificial intelligence model to perform parsing processing or case generation processing.

[0145] "Classification rules" refer to a set of logical conditions used to classify user information or other processed objects into different business or service categories, including combined rules based on keywords, attribute fields, model output results, and business strategies.

[0146] "Business differentiation" refers to the classification of information based on business functions or processing procedures, including different business categories such as consultation handling, technical support, sales processing, and operation and maintenance services.

[0147] "Service differentiation" refers to the classification of information based on the type or module of services provided, including online services, on-site services, self-service, and managed services.

[0148] "Judgment result" refers to the classification conclusion or label information made based on the application of case information and classification rules to determine which business category or service category the user information belongs to.

[0149] In one embodiment of the present invention, the server, terminal, and user collaborate to form the overall architecture for realizing the system functions of the present invention. The server is installed in a data center environment and employs a general-purpose server hardware platform, such as a rack-mounted computer with a multi-core central processing unit, a graphics processing unit (e.g., a general-purpose graphics accelerator card), and large-capacity main memory and solid-state storage. The server runs a general-purpose operating system (e.g., a UNIX-like operating system) and runs application server programs, a database management system, and a generative artificial intelligence model inference service interface on it. The terminal is configured as a portable or wearable display device, such as smart glasses or a portable tablet terminal, running a mobile operating system (e.g., a general-purpose mobile operating system) and communicating with the server via a wireless local area network or cellular mobile communication network. The user, as the end user of the system, wears the terminal in the actual working environment and interacts with the server through the terminal, thereby obtaining real-time application example information in real-world scenarios.

[0150] The server utilizes a general-purpose programming environment (such as a Python environment running on a UNIX-like operating system) to build information collection and processing programs. It further leverages software modules such as network access libraries (e.g., HTTP request libraries), web page parsing libraries (e.g., HTML parsing libraries), and data processing libraries (e.g., numerical computation and data analysis libraries) to automatically collect and preprocess publicly available information from the internet in a structured manner. The server establishes secure connections with external information sources via communication media. After receiving user information forwarded by terminals, it generates search strings based on pre-defined rules and sends search requests to multiple external information sources. At the hardware level, the server works in conjunction with network cards and switches to execute large-scale HTTP requests, thereby completing data scraping from multiple sites in a relatively short time.

[0151] After obtaining publicly available information, the server executes a unified parsing process for each webpage response. The server utilizes an HTML parsing library to construct a document object model (DOM) tree structure and filters nodes based on predefined "information extraction criteria." For example, the server extracts the text content of title nodes, body paragraph nodes, and table cell nodes as "text information," and stores URLs, timestamps, and source site identifiers as metadata in a relational database system. The server uses a unified data structure for storage, managing data in logical table formats such as "webpage record tables," "text fragment tables," and "organizational information tables." This structured storage method significantly reduces the probability of repeated crawling and parsing compared to traditional manual operations, achieving redundancy removal and index optimization within data management, thereby improving the speed of subsequent queries and retrieval.

[0152] When generating prompts for invoking generative AI models, the server employs a modular text concatenation strategy. The server combines user information, business conditions, business form information, and extracted text information according to a fixed template to form a composite input text containing task descriptions, output field specifications, and filtering conditions. The server can use one of the following prompt examples as a base template: "Based on the following webpage text, please extract the latest wireless communication technology cases related to logistics centers within the last 3 years."

[0153] For each case, please output: the type of technology used, the name of the organization involved, its country or region, whether it is a developing region, and the main effects (e.g., percentage increase in efficiency, decrease in error rate). Alternatively, you can use the following example prompt: "Customer business information: Large e-commerce logistics center with automated sorting lines and RFID tag system."

[0154] Please select the most relevant wireless communication technology case studies from the following text and generate an actionable application plan for this client, including implementation steps and precautions. By generating prompts with clearly defined field definitions and output format constraints, servers can significantly reduce the ambiguity and redundancy of generative AI model outputs, thereby improving the accuracy and efficiency of parsing and structured processing within the system. Compared to the traditional approach that relies on users freely writing query content, servers improve the standardization of model input through centrally managed prompt templates and an auto-fill mechanism, technically enhancing the stability of the entire natural language processing chain.

[0155] The generative artificial intelligence model selected by the server in this invention is preferably a large-scale language model based on a deep learning architecture. This model can employ a multi-layer self-attention neural network structure, including an embedding layer, a multi-layer encoder module, a feedforward network, and an output layer. When deploying the model, the server uses a graphics processing unit to perform floating-point matrix operations and uses a tensor computation library to implement linear transformations and weighted summations of the query vector, key vector, and value vector in the self-attention mechanism. During the model inference phase, the server encodes the input sequence composed of prompts and text information into a vector representation, captures semantic dependencies through a multi-head self-attention mechanism, and predicts the model's output text token by token at the output end. During the model training or retraining phase, the server can use supervised learning to minimize the cross-entropy loss function and update the model parameters through backpropagation and gradient descent optimization algorithms. In some implementations, the server can also employ data augmentation strategies, such as paraphrasing the same case text, random masking, and paragraph order perturbation, to enhance the model's robustness during training.

[0156] After generating case and organization information, the server performs performance indicator calculations and ranking on the organization information. The server retrieves numerical fields related to the organization, such as sales revenue, profit margin, and business scale, from publicly available information or external data services, and normalizes them using a unified indicator system. For example, the server can take the logarithm of sales revenue, then perform min-max normalization to generate standardized performance values ​​between 0 and 1; then, it uses weighted summation or ranking algorithms to rank the organizations. The server encapsulates these calculation steps as vectorized operations, using numerical calculation libraries to perform batch calculations on the central processing unit or graphics processing unit, thus maintaining high processing speed even with a large number of organizations. Compared to simple text keyword matching, the server, through explicit numerical performance indicators and ranking algorithms, transforms subjective experience into a repeatable calculation process, achieving a technical improvement in the process of constructing the organization list information.

[0157] When generating application case information, the server takes case information, organization list information, and business model information as input and calls a generative artificial intelligence model for secondary generation. The server explicitly indicates in the prompts the fields and step order that the model needs to output, such as "target description," "wireless communication technology used," "deployment location," "implementation steps," and "expected benefits." The server further performs structured parsing from the model's output text using rule matching or a lightweight parser, mapping key fields in natural language to a unified data structure. Internally, the server uses a tree-like or hierarchical data structure to represent the application case information so that the terminal can expand it by chapter during display. Through this combination of structured generation and parsing, the server not only automatically generates business application solutions but also significantly improves the flexibility of subsequent retrieval, filtering, and combination at the data structure level, thus demonstrating an improvement in computer data management technology.

[0158] In this invention, the terminal serves as the user interface and performs some preprocessing functions. The terminal establishes an encrypted communication connection with the server via a wireless communication module, receiving structured data containing application example information and organization list information. The terminal parses this data locally and maps it onto user interface components, such as list views, card views, and detail pages. In wearable display device scenarios, the terminal displays key information as overlaid text in the user's field of vision, while reserving interactive areas for users to select different application examples via gestures, buttons, or voice. The terminal maintains a limited cache structure locally, indexing recently received application example information to ensure some functionality remains available even under poor network conditions. By performing display format conversion and cache optimization on the terminal side, this invention reduces the amount of data transmitted per refresh on the terminal presentation link, thereby achieving a reduction in communication load.

[0159] The terminal collects the user's voice input through its built-in microphone and calls a local or remote voice recognition module to convert the audio signal into text commands. Based on the recognized text content, the terminal generates structured user information and sends it to the server. For example, a user can issue a verbal request through the terminal: "Please specify examples of the latest wireless communication technologies and propose a method for using logistics centers." The terminal recognizes this sentence as text and transmits it to the server along with its context. The server performs language recognition and translation on the text (if necessary), then converts it into standardized search strings and prompts, thereby driving the generative artificial intelligence model to perform new parsing and generation. Compared to traditional methods that rely solely on manual keyboard input for queries, this invention combines voice input with an automatic prompt generation mechanism, making the intention conversion during user interaction more automated and standardized. This shortens the intermediate steps between user needs and model invocation, reducing latency and improving accuracy in the overall interaction process.

[0160] In this invention, the user is positioned at the work site or in a business scenario, wearing or carrying a terminal and maintaining a communication connection with the server. The user expresses the type of information needed via voice or other input methods, such as requesting the latest case studies of wireless communication technology in a specific industry, or requesting optimization solutions for the current warehouse layout. The user views application case information generated by the server in real time on the terminal's display interface, including recommended deployment schemes, parameter setting examples, etc. While reading and implementing these suggestions, the user can provide feedback on the actual effects via voice, such as "The current solution has reduced picking time by approximately 20%, please further optimize network load during peak hours." After receiving such feedback, the server can record it as labeled training samples or evaluation data for further fine-tuning of the generative artificial intelligence model or adjusting the prompt statement template. Through this feedback-driven update mechanism, the server gradually improves the matching degree between the model output and the user's actual needs, thereby achieving continuous optimization of recommendation accuracy at the algorithm level.

[0161] When processing classification rules and judgment results, the server does not simply rely on keyword matching, but also uses generated application case information and organization list information as feature sources. The server can extract features such as industry tags, technology categories, and complexity levels from application cases and merge them with the intent field in the user information to form a high-dimensional feature vector. Based on this, the server applies a classification algorithm (such as a classification model based on gradient boosting decision trees or shallow neural networks) to generate a probability score for each business or service distinction, and selects the target classification with the highest probability as the judgment result. Compared to traditional static rule routing, this method implements data-driven dynamic routing logic internally, enabling more accurate allocation of user information to the most suitable business module. Therefore, this invention optimizes traditional computational logic in information routing path selection, improving the overall system's throughput and error distribution rate control.

[0162] The server's internal module division allows the entire processing flow to be divided into multiple functional units, including an external information acquisition module, a text cleaning module, a feature extraction module, a generative artificial intelligence invocation module, a performance indicator calculation module, an application case generation module, a classification and judgment module, and a terminal communication module. The server uses a unified message queue or data bus to transmit structured data objects between modules and leverages asynchronous processing mechanisms to execute multiple requests in parallel in a multi-core environment. Because most processing steps employ a streaming architecture, the server can start fetching and preprocessing the next batch of data while the previous batch is still undergoing model inference, thereby improving overall system throughput and reducing end-to-end response time. This modular and parallel processing approach constitutes one of the key factors in improving processing speed and computational efficiency in this invention.

[0163] In another implementation, the server can deploy the generative AI model on a separate inference service node, accessing the model via a remote procedure call (RPC) interface. During the RPC, the server segments the input sequence and assigns a unique identifier to each segment for matching and merging multiple concurrent responses. After receiving the model output, the server locally trims and reorders the results according to existing rules. For example, it prioritizes displaying application examples implemented by high-performing organizations based on performance indicators, or filters out solutions unsuitable for the current operating environment based on business context information. This post-processing strategy for model output prevents the generated content from becoming excessively long, structurally unclear, or deviating from business requirements in practical applications, thereby enhancing the controllability and interpretability of generative AI technology at the system level.

[0164] In another implementation, the server can pre-train offline on past user queries, usage examples, and user feedback data to build a domain-specific fine-tuned model. During this training process, the server employs supervised signals, assigning higher weights to high-quality output samples based on positive user feedback or implicit behaviors (such as multiple views or forwards). A sample weighting mechanism is introduced into the loss function, making the model more biased towards generating such high-rated outputs when updating parameters. Through this feedback-based learning mechanism, the server gradually forms local optima in the model parameter space, oriented towards a specific business domain. This enables the model to generate usage examples that are more relevant to the target domain during the online inference phase. This combination of training and inference, from the perspective of computer learning systems, represents a continuously evolving technical solution, rather than simply automating manually defined rules.

[0165] In summary, the server, through automatic collection and structured preprocessing of publicly available information, standardized invocation and multi-stage generation of generative artificial intelligence models, numerical calculation and ranking of organizational performance indicators, structured generation and parsing of application case information, data-driven optimization of classification rules and judgment results, and efficient communication and rendering coordination with the terminal, has achieved a technical solution that is applicable to specific business domains but has universal applicability within the computer. The terminal improves user interaction efficiency and system responsiveness through real-time display, voice input collection, and local caching. Users can obtain high-quality application case information in actual work scenarios through this system and participate in the system optimization process through feedback. Due to the close coupling of the above technical links, this invention not only improves information utilization efficiency at the business application level but also achieves quantifiable improvements in multiple aspects such as processing speed, data management accuracy, communication load control, and intelligent routing logic within computer technology, thus constituting a substantial improvement to existing computer and generative artificial intelligence technologies.

[0166] use Figure 12 The processing procedure is explained.

[0167] Step 1: The server receives user requests and parses user information.

[0168] Input: User information sent by the terminal via the network (including text content after speech recognition, optional service type information, language type, whether it needs to be organized and sorted, etc.).

[0169] Output: A structured user request object (containing standardized query intent fields, business condition fields, and filter condition fields).

[0170] The server reads JSON data from the HTTP request message, parses it to extract the user's query text (e.g., "Please list the latest wireless communication technology cases and propose their application solutions in logistics centers."), business type (e.g., logistics center), and whether "performance indicator ranking" is required. The server performs language detection and necessary translation on the query text, uniformly converting it into the target language used internally. Based on preset business rules, the server maps the query text to standardized intent tags and several parameter fields, such as "Target Technology: Wireless Communication Technology," "Application Scenario: Warehousing and Logistics," and "Regional Preference: Unrestricted," generating a user request object for internal use.

[0171] Step 2: The server generates an external information retrieval string and identifies the target information source.

[0172] Input: The user request object generated in step 1, the pre-configured business rules, and the list of external information sources.

[0173] Output: One or more search strings and a list of corresponding target URLs.

[0174] The server extracts keywords (such as "wireless communication technology," "logistics center," "case study," "developing region," etc.) from the user's request object and combines them into a search string according to a pre-defined template. The server then references a list of external information sources in its configuration file (such as general search engine entry URLs and industry website search interface URLs) to concatenate each search string into a complete request URL. The server also configures corresponding HTTP header fields, timeout periods, and retry policies for each URL, generating a list of target URLs required for subsequent network access.

[0175] Step 3: The server obtains publicly available information from external sources and performs text extraction and cleaning.

[0176] Input: The list of target URLs and the search string generated in step 2.

[0177] Output: A collection of extracted and cleaned text information and corresponding metadata (URL, timestamp, source, etc.).

[0178] The server uses an HTTP request library to access target URLs one by one and retrieve HTML response data. It then uses an HTML parsing library to build a document object model and extracts title text, body paragraphs, and technically relevant table content based on predefined information extraction criteria (such as body selectors, heading selectors, and table selectors). The server merges the extracted text into text information fragments and performs cleaning processes, including removing HTML tags, footnotes, ad tags, and redundant whitespace. Finally, the server appends metadata such as the source URL, fetch time, and language to each fragment, forming a structured record stored in an in-memory data structure or database, which can then be used as input candidates for subsequent calls to generative artificial intelligence models.

[0179] Step 4: The server constructs prompts for invoking generative artificial intelligence models.

[0180] Input: The user request object from step 1, the text information set from step 3, the pre-set business conditions, and the prompt statement template.

[0181] Output: A set of model input units containing prompt statements and corresponding text information fragments.

[0182] Based on the business context and target task in the user's request, the server selects a suitable template from the prompt template library, such as a "case extraction template" or a "solution generation template." The server then fills the template with the user's business context description (e.g., "large e-commerce logistics center with automated sorting lines and RFID tag systems") and filtering conditions (e.g., "last 3 years" or "developing regions"), generating the specific prompt text. According to the input length limit of the generative AI model, the server segments the text information set, combining each segment with a prompt text to form an input unit. The result is a combination of several "prompt text + text segments," each combination serving as a candidate input for a model call.

[0183] Step 5: The server invokes the generative artificial intelligence model to perform parsing and case generation processes.

[0184] Input: A set of model input units generated in step 4 (each unit contains a prompt statement and a corresponding text information fragment).

[0185] Output: Initial text results of case information and organizational information generated by the generative artificial intelligence model.

[0186] The server sends each input unit to the generative AI model service via a model inference interface. Internally, the server encodes the input text into a sequence of tokens and passes it to a multi-layer self-attention neural network deployed on a graphics processing unit. This network vectorizes the input, calculates the correlation between different tokens using a multi-head self-attention mechanism, performs a non-linear transformation of the hidden states using a feedforward network, and finally predicts the generated sequence token by token at the output layer. The server reads text results from the model's output, typically natural language paragraphs labeled with structured fields, such as "case name," "technology type," "organization name," "region," and "effect." The server parses these text results into intermediate structured objects according to a pre-defined format, distinguishing between case information and organizational information.

[0187] Step 6: The server calculates performance indicators based on organizational information and generates a list of organizations.

[0188] Input: The set of organizational information generated in step 5, and external or internal organizational data sources (such as sales, profits, etc.).

[0189] Output: A list of organizations with performance metrics and ranking results.

[0190] The server retrieves corresponding numerical fields (such as annual sales revenue and profit margin) from a pre-configured data source based on the organization's name or identifier. The server standardizes these values ​​(e.g., logarithmic transformation, normalization) and combines them into a single performance indicator according to configured weights. The server then sorts all organizations by performance indicator, generating a ranking for each organization. The server encapsulates the organization name, performance indicator value, and ranking result into an organization list information object, providing a basis for subsequent use case generation. Through vectorized operations and batch processing, the server reduces the overhead of calculating each organization individually, improving overall computational speed.

[0191] Step 7: The server generates application case information that matches the user's business model.

[0192] Input: The set of case information from step 5, the organization list information from step 6, and the business form information and business conditions from step 1.

[0193] Output: A structured collection of application examples.

[0194] The server combines the case information and the organization list information as context, and reconstructs a prompt statement, such as "Based on the following cases and the practices of high-performing organizations, generate three feasible solutions for the specified business model." The server sends this prompt statement, along with a summary of the case information and a description of the business model, to the generative AI model. Internally, the generative AI model performs joint encoding and semantic reasoning on this information, outputting solution text that includes implementation goals, technologies used, deployment locations, implementation steps, and expected benefits. The server parses this text according to preset tags, breaking down paragraphs in natural language into fields, constructing a hierarchical data structure (e.g., "Solution List" contains multiple "Step Lists"), forming a set of application case information. During this process, the server performs consistency checks on the output, such as verifying whether necessary fields are included and whether steps are sequentially numbered, to ensure the data structure is suitable for subsequent automated processing.

[0195] Step 8: The server generates display data for the terminal and sends it to the terminal.

[0196] Input: The set of application case information from step 7, the organization list information from step 6, the display layout template, and the terminal capability information (screen size, resolution, etc.).

[0197] Output: Display data adapted to the terminal (structured, compressed display data).

[0198] The server selects an appropriate display template (such as a concise mode or a detailed mode) based on the terminal's capabilities, organizing the title, key metrics, and main steps from the application examples into a structure suitable for small screens. For example, the server generates a "card object" for each application example, containing a title, technical tags, a summary of key effects, and links to the complete steps. The server combines these card objects with organizational ranking information to form display data, and then compresses and serializes the display data to reduce network transmission volume. Subsequently, the server pushes the display data to the corresponding terminal through an encrypted communication channel.

[0199] Step 9: The terminal receives the data to be displayed and presents it on the display device.

[0200] Input: Display data from the server.

[0201] Output: A visual interface on a portable or wearable display device.

[0202] The terminal receives display data sent by the server via a network interface, decompresses and deserializes the data, restoring it to internal objects. Based on its UI framework, the terminal maps application example cards to interface components, displaying the title bar and key metrics in a HUD overlay in wearable display scenarios, and expanding the detailed step list upon user request. The terminal performs pagination display based on a preset refresh strategy or user scrolling, and caches the latest batch of display data locally to continue providing reading functionality even when offline. The terminal uses a reasonable layout to reduce the number of elements loaded at once, thereby reducing rendering burden and improving interface responsiveness.

[0203] Step 10: Users browse application examples and generate operational needs through the terminal.

[0204] Input: The application example interface displayed in the terminal.

[0205] Output: The user's implicit or explicit requirements for subsequent queries or optimizations.

[0206] In real-world business scenarios, users wear or carry terminals and read application examples displayed on the screen, including recommended technology combinations, deployment location suggestions, and implementation steps. Users assess the feasibility of solutions based on the on-site situation, potentially generating new questions or optimization needs, such as requiring more detailed case studies for specific operational processes (e.g., peak-hour sorting). While this step primarily reflects user behavior, the system considers this behavior a prerequisite for subsequent interactions from an overall process perspective. The terminal can use implicit signals such as user dwell time and scrolling behavior to help determine the user's focus.

[0207] Step 11: The terminal collects user voice input and generates new user information.

[0208] Input: Voice commands or questions issued by the user near the terminal.

[0209] Output: User query content in text form and structured user information.

[0210] The terminal captures the user's voice signal through its built-in microphone and calls local or cloud-based speech recognition services to convert the audio stream into text. For example, if a user says, "Please give me another wireless communication technology solution suitable for nighttime work," the terminal obtains the corresponding text result. The terminal extracts keywords and intents (such as "nighttime work" and "wireless communication technology solution") from the text according to preset rules, constructing a new user information object. The terminal encapsulates this object into a new request message and sends it to the server, thus triggering a new round of processing. In this way, the terminal provides the server with continuous, high-frequency updates to user intents without relying on keyboard input.

[0211] Step 12: The server performs loop control based on the new user information and updates the results.

[0212] Input: The new user information object generated in step 11, historical usage case information, and user interaction records.

[0213] Output: Updated application examples and corresponding display data generated for the new requirements.

[0214] After receiving new user information, the server compares it with previously stored user interaction records to identify whether the current request is a refinement or supplement to the previous query. Based on this, the server adjusts the prompt template, for example, adding parameters such as "nighttime working conditions" or "low-light environment constraints" to the original template, and reconstructs the prompt. The server again selects relevant text information from external information sources or internal cache, and calls a generative artificial intelligence model for parsing and example generation. The server performs the same structured processing, performance indicator retrieval and sorting on the newly output example information as before, and generates new application example information and display data. The server returns the updated display data to the terminal, allowing users to see the optimized solution generated based on the latest intent in a short time, achieving a rapid "request-generation-feedback" cycle. Through this path, the server internally implements dynamic adjustments to prompt statements and data selection strategies, thereby improving the adaptability of the generated results to continuous user needs and overall processing efficiency.

[0215] 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.

[0216] 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."

[0217] With the continuous increase in data volume and business complexity, existing computer data processing systems face the following major technical problems when extracting, cleaning, sorting, and filtering large amounts of business information (such as financial information) according to user needs: First, existing systems typically rely on fixed-format query statements or predefined form parameters. Users must master structured query languages ​​or complex operation steps, making it impossible to directly and flexibly control the data processing flow through natural language. This results in low human-computer interaction efficiency and poor system adaptability to changes in requirements. Second, in existing systems, the application of generative artificial intelligence models is mostly limited to text generation or simple question-and-answer levels, and has not yet been tightly coupled with specific computer data processing processes such as database retrieval, data cleaning, sorting, and filtering. This makes it impossible to achieve automated generation and dynamic adjustment of query conditions, filtering rules, and output formats, leading to insufficient utilization of computing resources and a rigid overall processing flow. Third, when processing business information, especially financial information, existing technologies generally lack an integrated standardization mechanism for the internal data structure of computers. For example, systematic standardization of missing values, outliers, data types, units of measurement, and monetary units often requires the manual writing of a large number of scripts, leading to problems such as complex implementation, high maintenance costs, and high error rates during operation. Fourth, the existing system is unable to respond promptly to users' multi-round, iterative analysis needs. It cannot dynamically adjust the filtering conditions and sorting rules and automatically recalculate the results based on the same batch of standardized data through multiple rounds of natural language commands, resulting in repetitive calculations, slow system response, and impacting overall performance and user experience.

[0218] Therefore, it is necessary to provide a system that can deeply integrate the parsing capabilities of generative artificial intelligence models for natural language prompts with structured data retrieval, data regularization, and sorting and filtering within a computer. By introducing new control processes and data processing mechanisms on the server side, the system can improve the automation and flexibility of data processing, reduce reliance on manual programming, optimize the use of storage and computing resources, and thus substantially improve the performance and availability of computer technology in large-scale business information processing scenarios.

[0219] 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.

[0220] In this invention, the server includes a device for receiving natural language instruction information from a user via an input / output device; a device for sending the instruction information as a prompt to a generative artificial intelligence model and instructing the generative artificial intelligence model to generate structured condition information; a device for generating acquisition conditions for multiple target information in a storage device based on the structured condition information and extracting the target information through an information processing device; a device for performing regularization processing on the extracted target information, such as missing value processing, outlier removal, data type unification, and conversion of measurement and currency units, to generate regularized information; a device for performing sorting processing based on numerical attributes and extraction quantity restriction processing on the regularized information to generate a target element set with assigned order; and a control processing device for dynamically updating the structured condition information and the acquisition conditions according to the prompts added by the user and repeatedly performing extraction, regularization, and sorting processing. This enables an integrated, automated processing flow within the server, from natural language prompts to structured conditions, database retrieval, data organization, sorting, and filtering. The computer system can dynamically adjust its data processing logic based on multiple rounds of user-defined natural language commands without relying on fixed query statements or extensive manual scripts. This reduces human intervention, lowers implementation complexity and error rates, and improves storage access efficiency and computing resource utilization. Consequently, it significantly enhances the technical performance and interactivity of the computer system when processing large-scale business information, especially financial information.

[0221] A "system" refers to the overall computer implementation device composed of multiple functional components working together through hardware and software, used to receive user input, invoke generative artificial intelligence models, perform data acquisition and processing, and output results.

[0222] "Input / output device" refers to a device used to exchange information between humans and machines, including an input interface for receiving user input and an output interface for presenting processing results to the user. It can consist of a display, keyboard, pointing device, touch screen or communication interface, etc.

[0223] "User" refers to the human-computer interaction participant who provides instruction information to the system and receives output information from the system through input / output devices.

[0224] "Natural language instructions" refers to text or voice content expressed by users in natural language to instruct the system to perform operations such as data collection, filtering, sorting, and exporting.

[0225] "Prompt statements" refer to natural language text extracted from natural language instructions or directly input by the user as input to a generative artificial intelligence model, used to trigger the automatic generation of conditions, rules, or operational steps.

[0226] "Generative AI models" refer to AI models that can perform natural language understanding and generation tasks based on input prompts. These include models that use machine learning and deep learning techniques to perform semantic parsing of text and output structured information or natural language content.

[0227] "Structured conditional information" refers to a set of conditions generated by a generative artificial intelligence model based on prompts and represented in a predetermined structure. These conditions include parameters such as business domain, time range, and evaluation items, which can be directly parsed and applied by computer programs.

[0228] "Business domain" refers to the scope of application or industry category used in business information classification to define the business scope to which the target information belongs.

[0229] "Time range" refers to the conditions used in data processing to limit the time interval to which the target information belongs, including year, accounting period, start and end dates, etc.

[0230] "Evaluation items" refer to attribute items in target information used as the basis for analysis or ranking, including numerical indicators and categorical indicators, such as revenue, profit, assets, liabilities, etc.

[0231] "Information storage device" refers to data storage resources used to store multiple target information, including database systems, file systems, or other storage media accessible by a computer.

[0232] "Target information" refers to data records stored in information storage devices that can be extracted based on structured conditions, including business information, financial information, or other structured data.

[0233] "Retrieval conditions" refers to a specific set of conditions generated based on structured condition information, used to retrieve target information from an information storage device. These conditions can correspond to query statements or filtering rules.

[0234] "Information management function" refers to the set of functions performed by information processing devices for accessing, retrieving, updating and managing stored data, including operations such as querying, filtering, sorting and writing.

[0235] "Information processing device" refers to computer hardware used to perform information management functions and data processing programs, and the software running on it, including processor, memory and application programs.

[0236] "Information processing program" refers to a software program or script that runs on an information processing device and is used to organize, sort, and filter extracted target information.

[0237] "Missing value handling" refers to the processing operations performed during data normalization to complete, delete, or replace missing data items in the target information.

[0238] "Outlier removal" refers to the process of identifying and removing abnormal data records that do not conform to a reasonable range or statistical distribution during the data normalization process, according to predetermined rules.

[0239] "Data type unification" refers to the process of converting different data types of the same attribute in the target information during data standardization so that they are represented by a unified data type.

[0240] "Unit conversion" refers to the process of converting values ​​from different units of measurement into a unified unit of measurement during data normalization.

[0241] "Currency unit conversion" refers to the process of converting amounts in different currency units into a unified currency unit during data normalization, based on a preset exchange rate or conversion rules.

[0242] "Regulation processing" refers to a series of data cleaning and standardization operations performed on the extracted target information, such as handling missing values, removing outliers, unifying data types, and converting units.

[0243] "Regulated information" refers to a set of target information that has been processed, where missing and outlier values ​​have been removed, and the data type and unit are consistent.

[0244] "Numerical attributes" refer to attributes in target information that are represented in numerical form and are used for comparison, sorting, and calculation, including amounts, quantities, ratios, etc.

[0245] "Sorting" refers to the process of arranging records in a structured information according to a predetermined order based on numerical attributes or other indicators.

[0246] "Extraction quantity limit processing" refers to the processing operation of selecting a portion of records from the sorted results after sorting the regular information, based on a predetermined quantity limit or range.

[0247] "Target element" refers to the basic unit that constitutes a single record or entity in the target information, including the organization and its corresponding attribute items.

[0248] "The set of target elements assigned an order" refers to a set of multiple target elements arranged in a specific order after sorting and extraction quantity restriction processing.

[0249] "Specific information" refers to the set of information that is output and consists of a set of target elements that have been assigned a specific order.

[0250] "Business information" refers to a set of data that describes the status or results of business activities, including various business-related data such as financial information.

[0251] "Financial information" refers to a set of data that reflects an organization's economic status and operating results, including indicators such as revenue, profit, assets, and liabilities.

[0252] "Organization" refers to the entity that serves as the subject of financial or business information, including enterprises, institutions, or other economic entities.

[0253] "Sales revenue" refers to the amount of income an organization receives by selling goods or providing services within a certain time frame.

[0254] "Control and processing device" refers to the processing unit used to coordinate and control the execution order and parameter updates of various processing steps such as input and output, generative artificial intelligence model invocation, data extraction, normalization, and sorting and filtering.

[0255] "Additional prompts" refer to prompts that users can enter after the initial data processing flow to adjust filter criteria, sorting rules, or output format.

[0256] "Condition parsing" refers to the process by which generative artificial intelligence models perform semantic analysis on prompt statements to extract structured information such as filtering conditions, sorting conditions, and output format conditions.

[0257] "Filtering conditions" refer to the set of logical conditions generated based on the results of condition parsing, used to select data records that meet specific constraints from regularized information.

[0258] "Sort criteria" refers to the set of rules used to specify the sort key and sort order used in the sorting process.

[0259] "Output format conditions" refers to a set of conditions used to specify the output result format, field combination, report structure, or file format.

[0260] "Data information format for reports" refers to the output data format organized according to a predetermined report structure or file format, including data structures used for report display or exported files.

[0261] In a preferred embodiment of the present invention, the server serves as the main computing node and the terminal serves as the human-computer interaction node. The user sends natural language prompts to the server through the terminal. The server then calls a generative artificial intelligence model and combines it with database management software and data analysis library to perform a series of specific data processing and data operations on general-purpose computer hardware, thereby achieving efficient extraction, organization and sorting of business information, especially financial information.

[0262] At the hardware level, servers can utilize general-purpose server equipment, such as rack-mount computers with multi-core central processing units, main memory, solid-state storage, and network interfaces. At the software level, servers can run general-purpose operating systems, such as Unix-like operating systems; on top of this operating system, the server can deploy database management system software, such as relational database management systems; scripting language runtime environments, such as interpreted programming language runtime environments; and data analysis libraries, such as libraries for processing data table structures. Servers can also deploy web application frameworks to implement communication interfaces with terminals.

[0263] The terminal can be a desktop computer, a laptop computer, or a mobile communication terminal. At the hardware level, the terminal includes a display device, an input device, and a network communication module; at the software level, it includes a web browser or a dedicated client application. Users input natural language text through the terminal's user interface and view data lists and reports returned by the server.

[0264] After receiving a natural language prompt from a user via a terminal, the server stores the prompt as a text string in memory and encapsulates this string as input to a generative AI model. This generative AI model can be deployed on local acceleration hardware (such as a graphics processing unit) or accessed as a service over a network. The server can invoke a sequence-to-sequence neural network model based on a self-attention mechanism. This model can employ a multi-layer encoder and multi-layer decoder structure, with each layer including a multi-head self-attention sublayer and a feedforward network sublayer. The server can use word embeddings and positional encoding in the model to map characters or tokens to real-valued vectors, and use scaled dot product attention in attention computation to improve the efficiency and accuracy of parsing long sequence instructions.

[0265] During model training, the server can use a large number of historical business query statements and corresponding structured condition pairs as training data. The server can use the cross-entropy loss function to evaluate the difference between the conditionally labeled sequence and the labeled sequence output by the model, and update the model weights through backpropagation. During training, the server can employ variants of batch gradient descent, such as adaptive learning rate optimization algorithms, to accelerate convergence and reduce oscillations during training. When constructing training data, the server can perform data augmentation, such as using different natural language expressions for the same condition, synonym replacement, and word order perturbation, to improve the model's robustness to various prompt statement formats.

[0266] During the inference phase, the server inputs the user's prompts into a trained generative AI model. Based on the learned parameters, the model semantically decomposes the prompts, extracting features such as business domain, time frame, and evaluation items, and outputs structured conditional information. After receiving the model's output, the server maps these conditions into an internal conditional data structure, such as a collection of conditional objects, where each object includes a field name, comparison operator, and comparison value. The server then constructs a database query request based on this internal conditional data structure, avoiding the need for users to directly write query language, thus automatically completing the conversion from natural language to structured queries within the computer.

[0267] When extracting target information, the server generates and executes queries through the database management system. The server can use index structures, such as creating B-tree indexes on the industry, year, and amount fields of the financial information table. When constructing queries, the server prioritizes using indexed fields to reduce the probability of full table scans, thereby reducing disk accesses and improving data retrieval speed. After executing the query, the server loads the results into a row-column data structure, such as a DataFrame Object. This data structure is organized in memory in a columnar storage format, with each column maintaining the same data type to improve cache hit rate for batch operations and the efficiency of vectorized operations.

[0268] During the normalization process, the server performs a series of data cleaning and standardization operations on the data frame object. When handling missing values, the server can choose different strategies based on the field type. For example, it can use the median or mean to impute monetary values, the mode to impute categorical values, or delete the entire row if a key field is missing. When removing outliers, the server can calculate the quantiles or standard deviation of each numerical field using statistical methods and determine whether it is an outlier based on a preset threshold. When an outlier record is detected, the server can set a flag or directly exclude these records to reduce the impact of extreme values ​​on subsequent calculations.

[0269] When unifying data types, the server checks the internal representation type of each field and parses and converts strings that cannot participate in numerical calculations. For example, it parses text containing currency symbols and thousands separators into floating-point numbers. When converting units of measurement, the server can maintain a unit conversion table to convert quantities expressed in different units to a base unit. When converting currency units, the server can maintain an exchange rate table to scale amount fields proportionally based on a specified base date or range. Through these standardization processes, the server transforms the raw data into standardized information that is consistent in numerical calculations.

[0270] When sorting organized information, the server can generate a sort key vector based on numerical attributes and use an optimized sorting algorithm (such as improved quicksort or mergesort) to sort the records in the data frame. The server can confine the sorting operation to memory blocks and employ an external sorting strategy when the data volume is too large, writing intermediate results to external storage and merging them using a merge process to control memory usage. After sorting, the server can extract a predetermined number of records to form a set of target elements with assigned order, such as a set of organizations where operating revenue is prioritized.

[0271] When displaying specific information returned by the server, the terminal can parse the JSON data into a table structure and display it as a table or chart in the graphical user interface. The terminal can also allow users to further rearrange the display order or hide certain columns locally using controls such as sorting buttons and filter input boxes, thereby improving the efficiency of visual interaction.

[0272] When using this system, users can input various prompts on the terminal, such as "Please collect the financial data of major companies in China's IT industry for the past year, including operating revenue, net profit, total assets, and liabilities," or "Based on the data in the current database, please list the top 10 companies in the IT industry by operating revenue and provide their operating revenue and net profit." Users can also input more complex combinations of conditions, such as "Among A-share listed companies in China, find companies with an operating revenue growth rate greater than 20% in the past year and positive net profit, and rank the top 30 in descending order of operating revenue." After receiving these prompts, the server uses a generative artificial intelligence model to parse the semantics and automatically generates corresponding structured conditions and calculation rules.

[0273] When processing conditions such as "revenue growth rate greater than 20%", the server internally constructs new derived field calculation rules based on the model parsing results. Specifically, it calculates the growth rate field based on historical revenue data from multiple periods. The server adds a new column to the data frame object, uses vectorized operations to calculate the growth rate value for each record, and then uses this column in filtering and sorting operations. This natural language-triggered derived feature calculation differs from traditional filtering based solely on static fields. The server can dynamically construct new computational graphs, thereby automating the evaluation of complex financial indicators.

[0274] The server employs a modular data processing pipeline structure throughout the data flow. The generative AI model parsing module outputs condition objects, which are then passed to the query module. The query module outputs raw data frame objects, the normalization module outputs normalization information, and the sorting and filtering module outputs specific information. Each module is connected through a clearly defined data structure. The server can independently optimize each module; for example, multi-threading or single-instruction multi-data instructions can be used in the normalization module to accelerate numerical conversion, while a lazy evaluation strategy can be used in the sorting module to sort only the fields that need to be displayed, thereby further reducing computational redundancy.

[0275] The server employs an unconventional processing path—from natural language to structured conditions to vectorized data processing—enabling the computer system to automatically select efficient data structures and algorithms, rather than simply simulating manual filtering and sorting. Internally, the server utilizes columnar data structures and batch processing, combined with indexing and external sorting techniques, enabling the system to maintain high throughput and low latency even on large datasets. Compared to traditional systems based on fixed form parameters and static query statements, this system eliminates the need for manual script rewriting when handling multi-round iterative queries and dynamic condition combinations, thus achieving a technological improvement in computational resource utilization and response speed.

[0276] In another implementation, the server can restrict the output of the generative AI model to a specific function call format. This means the server requires the model to directly generate intermediate language expressions representing internal conditions, such as abstract syntax trees containing field names, comparison operators, and logical joins. Upon receiving these intermediate expressions, the server can directly construct and optimize the query plan, for example, by rearranging condition order to prioritize indexed fields, merging redundant conditions, and performing predicate pushdown. This structured output offers greater machine readability and optimizability compared to simply generating natural language descriptions, enabling the server to perform fine-grained performance optimizations at the execution plan level, thereby further reducing access latency.

[0277] In another implementation, the server can assess the confidence level of the parsing results from the generative AI model. When the confidence level of certain conditions is low, the server can generate an interactive confirmation interface on the terminal, allowing the user to select whether the parsed conditions contain the desired field or threshold. The server can update the model training samples based on the user's confirmation, thereby improving the parsing accuracy in subsequent iterations. This human-machine collaborative feedback mechanism, combined with the server's online or offline incremental training capabilities, can continuously improve the model's performance in the specific application scenario while maintaining stable system operation.

[0278] In this invention, the server does not merely automate business processes in a formal sense, but introduces new technologies at the level of internal computer data structures, algorithm paths, and execution plan generation. It maps natural language prompts into high-level conditional descriptions and derived feature definitions using a generative artificial intelligence model, and then executes these through vectorized regularization and sorting algorithms oriented towards columnar data structures, ultimately generating technically optimized query results. This top-down, interconnected design causally leads to improvements in processing accuracy and speed: on the one hand, regularization reduces statistical errors caused by noisy data; on the other hand, the combination of conditional parsing and query optimization reduces disk I / O and network communication, thereby improving overall system throughput and responsiveness.

[0279] In all embodiments of this invention, neither the terminal nor the user needs to understand the underlying data table structure and algorithm details; they only need to express their needs in the form of prompts. However, the server internally transforms these natural language requirements into hardware resource-friendly execution paths through specific model structures, data structures, and processing pipelines, thereby achieving substantial improvements to computer technology itself—including storage management, computation execution, and network interaction.

[0280] use Figure 13 The processing procedure is explained.

[0281] Step 1: The user enters a prompt message in the terminal and sends it to the server.

[0282] The user types natural language prompts into the terminal's input interface, such as "Please collect the financial data of major Chinese IT companies for the past year, including operating revenue, net profit, total assets, and liabilities," or "Based on the data in the current database, please list the top 10 IT companies by operating revenue and provide their operating revenue and net profit." After the user clicks the send button on the terminal, the terminal encapsulates this prompt, along with metadata such as user identifier and session identifier, into a request message and sends it to the server via network protocol. During this process, the terminal encodes the text into a Unicode encoding format and appends authentication information to the message header. The input for this step is the natural language prompt entered by the user on the terminal, and the output is the request data packet sent from the terminal to the server.

[0283] Step 2: The server receives the prompt and invokes a generative artificial intelligence model to perform condition parsing.

[0284] After receiving a request data packet from the terminal via the network interface, the server parses the prompt statement from the message body and stores it as a text string in memory. The server then uses this text string as input and passes it to the inference interface of the locally or remotely deployed generative AI model. When invoking the model, the server constructs a request structure containing the prompt statement and task type identifier, and sends it to the model service via a high-performance communication protocol. Internally, the generative AI model uses a multi-layer self-attention neural network to encode the prompt statement, extracting features such as business domain, time range, evaluation items, and screening thresholds from the instructions to generate structured conditional information, such as industry fields, year fields, and a list of indicators to be calculated. After receiving the structured conditional information returned by the model, the server converts it into an internal set of conditional objects. The input to this step is the prompt statement text, and the output is the server's internal structured conditional information data structure.

[0285] Step 3: The server generates database query conditions based on structured condition information and extracts target information.

[0286] The server takes structured condition information as input and constructs query conditions for a relational database based on the industry, region, time range, and indicator fields it contains. The server maps the condition objects to filter expressions in the query statement, prioritizing the use of existing index fields to reduce scan costs. The server then sends a query request to the database management system via the database connection interface, and the database returns a set of records that meet the conditions. The server loads the record set into a columnar data structure (such as a data frame), where each column corresponds to an attribute (such as company name, industry, fiscal year, operating revenue, net profit, etc.). During this process, the server performs preliminary conversion on the raw data type returned by the database to adapt it to the format requirements of the data frame. The input to this step is structured condition information, and the output is a target information data frame stored in memory.

[0287] Step 4: The server performs missing value processing and outlier removal on the target information.

[0288] The server takes the target information data frame as input and performs data quality checks on each field. First, the server scans key numerical fields (such as operating revenue, net profit, total assets, and liabilities), detecting missing values ​​and processing them according to pre-configured strategies: for non-key fields, statistical imputation is used; for key fields, entire rows are deleted. The server then performs statistical analysis on the numerical distribution, calculating the mean, standard deviation, or quantiles for each field, and identifying outlier records based on set thresholds. The server marks or removes these outlier records to reduce the impact of extreme data on subsequent calculations. The input for this step is the original target information data frame, and the output is an intermediate data frame that has completed missing value processing and outlier removal.

[0289] Step 5: The server performs data type unification and unit conversion on intermediate data frames to generate standardized information.

[0290] The server takes the cleaned intermediate data frame as input, checks the data type of each column, parses and converts numeric fields stored in text format, unifying them into a data type that can participate in numerical calculations. Next, the server normalizes fields using different units of measurement according to a pre-stored unit conversion table, converting them to a unified standard unit. For currency amount fields, the server converts amounts in different currency units to a unified base currency according to an exchange rate table. After completing these conversions, the server generates a standardized information data frame that is unified in numerical scale and data type, laying the foundation for subsequent sorting and derived index calculations. The input to this step is the intermediate data frame after handling missing and outlier values, and the output is the standardized information data frame.

[0291] Step 6: The server calculates derived indicators and generates sort keys based on the regularized information.

[0292] The server takes a structured information data frame as input and calculates necessary derived indicators, such as revenue growth rate, net profit margin, or debt-to-equity ratio, based on evaluation items and threshold conditions parsed by a generative artificial intelligence model. The server adds a new column to the data frame and calculates the corresponding derived value for each record through vectorized operations. Subsequently, the server constructs a sort key vector based on the sorting criteria implicitly or explicitly specified in the user request (such as sorting by revenue, revenue growth rate, or net profit margin). The server can combine multiple sorting fields to form a composite sort key to meet complex sorting requirements. The input to this step is a structured information data frame and calculation rules extracted from structured conditions; the output is an enhanced data frame containing derived indicator columns and the sort key definition.

[0293] Step 7: The server performs sorting and quantity restrictions on the enhanced data frames, generating a set of target features that are assigned an order.

[0294] The server takes an enhanced data frame and its sort key as input, and calls a sorting algorithm to arrange the data in ascending or descending order according to the specified numerical attributes. When the data volume is large, the server can use an external sorting strategy to divide the data into multiple blocks for local sorting, and then merge them into a globally ordered sequence through a merge process, thereby controlling memory usage. After the server completes the sorting, based on the quantity limit in the user request (e.g., top 10, top 20, top 50), it extracts the first few records of the sorted results to form a target element set with an assigned order, such as the set of the top 10 organizations in terms of operating revenue. The input of this step is an enhanced data frame and a sort key, and the output is a set of target elements arranged in order and with a limited quantity.

[0295] Step 8: The server encapsulates the target set of elements into an output structure and sends it to the terminal.

[0296] The server takes a set of target elements, ordered as input, selects the fields to be returned to the user, such as organization name, industry category, fiscal year, various financial indicators, and derived indicators. The server organizes these fields into a hierarchical data structure, adds metadata (such as generation time, total number of records, and filter summary), and serializes it into a transmission format. The server then sends this serialized result as a response message to the terminal via a network protocol. The input to this step is the set of target elements, and the output is the response data transmitted from the server to the terminal.

[0297] Step 9: The terminal receives and displays the results, and the user can enter additional prompts as needed.

[0298] After receiving the server's response at the network interface, the terminal parses the message body into an internal data structure and displays the ordered set of target elements in a table or chart format on the user interface. The terminal allows users to make minor adjustments to the display order or visible columns locally, but the core sorting and filtering logic is still executed by the server. If the user needs to further narrow down the scope or change the sorting criteria while viewing the results, they can enter additional prompts on the terminal interface, such as "Filter these companies for those with a net profit margin greater than 15%, and list the top 10 by net profit margin." The terminal resends this additional prompt along with the current session identifier to the server. The input for this step is the result data returned by the server and the user's interaction; the output is a new request for additional prompts.

[0299] Step 10: The server performs condition parsing based on additional prompts and repeatedly performs filtering and sorting on the regularized information.

[0300] The server takes the additional prompts and existing session state as input, and calls the generative AI model again to parse the conditions, obtaining new filtering conditions, sorting conditions, and output specifications. Instead of re-extracting the original data from the database, the server directly applies the new conditions to the previously generated regularized or enhanced data frames, thus avoiding redundant I / O overhead. The server constructs Boolean filtering expressions based on the parsed filtering conditions, performs row selection on the data frames, and then recalculates or selects the sort key according to the new sorting conditions, sorting and truncating the filtered records to form an updated target feature set. Finally, the server encapsulates the updated results and sends them to the terminal. The input to this step is the additional prompts and existing regularized information; the output is a regenerated, ordered target feature set based on the new conditions.

[0301] 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".

[0302] In existing information processing systems, user-input text or request content is typically used only as the basis for rule matching or simple classification, with fixed logic assigning requests to pre-defined processing modules or departments. These systems suffer from the following technical problems: First, they struggle to integrate structured and unstructured data from diverse, heterogeneous data sources in a timely manner. They cannot uniformly model and extract features from user attributes, environmental information, business requirements, and external database information within the same processing flow, resulting in insufficient adaptability of the generated processing results to specific scenarios. Second, even when generative AI models are introduced, existing systems typically rely on manually written static prompts for invocation. They cannot dynamically generate prompts based on real-time input data and context, making it difficult for generative AI models to fully utilize multi-dimensional features, thus limiting the quality and relevance of the generated use cases or solutions. Third, traditional workflow systems and rule engines are often isolated from intelligent generation modules. These systems lack a mechanism to automatically link use cases or solutions output by generative AI models with internal rule sets, failing to complete the closed-loop processing of "intelligent generation—rule mapping—automatic assignment—external device control" within a unified architecture, leading to significant manual intervention and repetitive configuration. Fourth, existing systems often ignore feedback data generated on the terminal and user sides, failing to incorporate user evaluation information, behavioral feedback information, and execution result information as training data back into the same data pipeline. This prevents the adaptive updating of generative AI models and rule sets during operation, leading to performance degradation after long-term operation and difficulty in coping with environmental and business changes. Fifth, in scenarios involving real-time control or resource allocation (such as network configuration and production line parameter adjustments), existing systems typically employ fixed strategies or static priority control, failing to prioritize requests and control commands based on dynamic factors such as user emotional state and use case urgency. Consequently, they suffer from significant latency and uneven response in computer resource scheduling and network control. This invention aims to provide a new system architecture and its processing mechanism, improving data integration, automatic generation of prompts, invocation of generative AI models, rule-driven automatic dispatch, and closed-loop feedback update capabilities from a computer technology perspective. This, in turn, enhances overall information processing efficiency, intelligence level, and adaptability to complex scenarios.

[0303] 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.

[0304] In this invention, the server includes a receiving unit for receiving attribute information, environmental information, and demand information from the user via an input / output device; a data processing unit for storing the received information, as well as structured and unstructured data obtained from external information sources, into a storage unit and integrating, normalizing, and extracting features from the data to generate data to be processed; a prompt generation unit for automatically generating prompt statements for inputting into a generative artificial intelligence model based on the data to be processed and conditional information corresponding to the user's purpose, and issuing an instruction to the generative artificial intelligence model to input the prompt statements to generate use cases or solutions; a classification unit for allocating and classifying the user's information to processing entities within their organization based on classification information and meta-information contained in the use cases or solutions output from the generative artificial intelligence model and a rule set stored in the storage unit; a control unit for calculating control values ​​or setpoints based on the use cases or solutions using observation data obtained from external devices and sending the control values ​​or setpoints to a terminal device via a communication network; and an update unit for receiving user evaluation information or feedback information from the terminal device and importing it again as the data to be processed to update the generative artificial intelligence model or the rule set. This allows for a closed-loop processing flow within the same computer system, encompassing "multi-source data integration—feature extraction—automatic generation of prompt statements—invocation of generative artificial intelligence models—rule-driven automatic assignment—control distribution to external devices—feedback and relearning." This enables the server to efficiently generate and execute intelligent use cases or solutions in complex scenarios using both hardware and software resources, significantly improving the computer's technical performance in data processing, resource scheduling, and adaptive updates.

[0305] "Input / output device" refers to a hardware or software interface used for data interaction between a user and an information processing device, including but not limited to a display, keyboard, mouse, touch screen, voice input device, camera device, and network-based graphical user interface or application interface.

[0306] "User" refers to the entity that provides attribute information, environmental information, and requirement information to the system through input / output devices and receives the use cases or schemes and control results output by the system. It can be a natural person or an operator who performs the operation on behalf of an organization.

[0307] "Attribute information" refers to data used to describe the static characteristics of users or target objects, including but not limited to basic characteristic information such as industry category, organization size, geographical distribution, business type, and role category.

[0308] "Environmental information" refers to data used to describe the state of a target scenario or operating environment, including but not limited to dynamic or semi-dynamic contextual information such as network conditions, device status, geographical location, time information, resource availability, and socio-economic conditions.

[0309] "Requirement information" refers to the content of a user's request to the system for a specific task or goal, including but not limited to problem description, goal constraints, expected output format, and priority requirements.

[0310] "External information sources" refers to various data providers or information systems located outside the system that provide data to the server, including but not limited to network databases, file storage systems, sensor networks, and third-party service interfaces.

[0311] "Structured data" refers to data with a fixed pattern or clearly defined field structure that can be directly managed by relational databases or tables, such as tabular data, relational database records, and record data with clearly defined fields.

[0312] "Unstructured data" refers to data that does not have a fixed field structure or pattern and is difficult to represent directly in a tabular format, such as natural language text, image data, audio data, video data, and free-format documents.

[0313] "Information processing device" refers to a data processing device that includes a processor, memory and communication interface, and is used to execute program code to complete data reception, storage, processing and output. It can be a single computer, a server cluster or a distributed processing system.

[0314] "Storage unit" refers to a data storage resource located inside or in communication with an information processing device, used to store structured data, unstructured data, rule sets, model parameters, and intermediate processing results, including volatile memory and non-volatile memory.

[0315] The “data processing unit” refers to the functional module executed by the processor, which integrates, normalizes, cleans, extracts features, and transforms received information and data from external information sources to generate data to be processed that can be used by subsequent models and rules.

[0316] "Integration" refers to the process of loading, aligning, and associating data from different sources and in different formats, so that they can be analyzed and processed under a unified data model or data view.

[0317] "Normalization" refers to the process of converting numerical or categorical data into a standard range or uniform expression form suitable for unified processing and modeling through mathematical transformations such as scaling, standardization, or encoding.

[0318] Feature extraction refers to the process of automatically or semi-automatically selecting and generating features that can represent the key attributes of the data from the original data, in order to reduce the dimensionality of the data and improve the performance of subsequent models or rule-based reasoning.

[0319] "Data to be processed" refers to a set of data that has been preprocessed through steps such as integration, normalization, and feature extraction, and can be directly input into generative artificial intelligence models or rule engines for reasoning and decision-making.

[0320] "Conditional information" refers to the parameters and contextual information that correspond to the user's purpose or system task objectives and are used to constrain or guide the results generated by the generative artificial intelligence model, including the target type, constraints, preference settings, and applicable scenario descriptions.

[0321] "Generative artificial intelligence models" refer to artificial intelligence models trained through machine learning or deep learning methods that can automatically generate text, solutions, use cases, or other information content based on input prompts.

[0322] "Prompt statements" refer to structured or semi-structured text instructions that are constructed to guide generative artificial intelligence models to produce expected outputs, including input context, constraints, and output requirements.

[0323] "Use case" refers to a contextualized solution generated by a generative artificial intelligence model based on the data to be processed and prompts, which describes the specific usage, process and effect of a technology, resource or service in a specific scenario.

[0324] A “solution” refers to a comprehensive set of processing plans or strategies output by a generative artificial intelligence model for a specific problem or goal, including the arrangement of steps, resource allocation, and description of expected results.

[0325] The "prompt generation unit" refers to a functional module implemented in the server, which is used to automatically construct prompt statements based on the data to be processed and conditional information, and issue instructions to the generative artificial intelligence model to use the prompt statements for generation.

[0326] "Classification information" refers to category tags or labels included in a use case or scenario, used to indicate the category, application area, service type, or business line to which the use case or scenario belongs.

[0327] "Meta-information" refers to descriptive additional information related to use cases or solutions, including but not limited to auxiliary information such as generation time, scope of application, confidence level, priority, and dependencies.

[0328] A "rule set" refers to a collection of multiple rules that are pre-defined in a storage unit and can be invoked by a processing unit. These rules are used to map user information and use cases or solutions to specific processing entities, control strategies, or processing flows.

[0329] A “classification unit” refers to a functional module executed on the server, which is used to automatically match and allocate user information to departments, service modules or processing queues within the organization based on classification information, metadata and rule sets.

[0330] "Processing entity" refers to a functional entity within an organization responsible for processing assigned information, including but not limited to business departments, service modules, processing queues, or automated execution modules.

[0331] "External devices" refer to various types of equipment that communicate with the server to collect environmental data, execute control commands, or provide services, including sensors, actuators, terminal devices, production equipment, and network nodes.

[0332] "Observation data" refers to various types of measurement data collected by external devices during operation and sent to the server, reflecting the status of equipment, environmental conditions, or business operations.

[0333] "Control value" refers to a parameter value calculated by the server based on use cases or schemes and observation data, used to directly drive or adjust the operating status of external devices.

[0334] "Settings" refer to target values ​​calculated and issued by the server for configuring operating parameters, thresholds, or policies of external devices or terminal devices.

[0335] "Terminal device" refers to user-side or field-side equipment that communicates with the server, used to receive control values ​​or set values, execute control commands, and return feedback information to the server, including mobile terminals, embedded devices, industrial controllers, or other subordinate devices.

[0336] A “control unit” refers to a functional module that runs on a server and is used to calculate control values ​​or setpoints based on use cases or schemes and observation data, and send them to the terminal device through a communication network to control external devices.

[0337] "User evaluation information" refers to quantitative or qualitative evaluation data given by users or field operators regarding the use cases, solutions, or execution results output by the system, including ratings, tags, comment text, and selection results.

[0338] "Feedback information" refers to data generated by the terminal device or user after executing a use case or plan and returned to the server, reflecting the execution effect, abnormal situation, or user reaction.

[0339] The "update unit" refers to a functional module executed in the server, which receives user evaluation information or feedback information, incorporates it into the data to be processed, and adjusts or retrains the parameters or rule set of the generative artificial intelligence model accordingly.

[0340] "Emotion inference model" refers to a machine learning or deep learning model that outputs a user's emotional state label or emotional intensity based on voice data, image data, and character data.

[0341] "Emotional information" refers to labeled data or vector representations output by an emotion inference model, used to represent the category and intensity of a user's emotions at a specific moment.

[0342] The "emotion processing unit" refers to a functional module that executes on the server to acquire and process emotional information, and dynamically adjust the content of prompt statements and the conditions for generating use cases or solutions based on the emotional information.

[0343] "Urgency information" refers to descriptive or numerical information used to reflect the urgency, risk level, or response time requirements of a task in a use case or solution.

[0344] "Priority score" refers to a numerical indicator calculated by the server based on sentiment and urgency information, used to rank the order of user information processing or the order of application of control commands.

[0345] The "priority control unit" refers to a functional module that runs on the server and is used to calculate priority scores and optimize and schedule the processing order of user information and the application order of control values ​​or settings based on these scores.

[0346] In this embodiment of the invention, the server, terminal, and user collaborate to form a technical system that automatically generates use cases or solutions using a generative artificial intelligence model and links a rule engine with external device control. The following describes the specific implementation methods of this invention from aspects such as system composition, data structure, model structure, learning methods, and technical effects.

[0347] I. Example of System Hardware and Software Composition Servers can use general-purpose computing devices, such as information processing devices equipped with multi-core central processing units, graphics processing units, large-capacity random access memory, and persistent storage devices. Servers can run Unix-like operating systems and install database management programs, web service programs, and machine learning frameworks.

[0348] The server can use the following software components to implement the functions of this invention: The server can use Python as the primary programming language and Django or similar web frameworks to implement application logic and interfaces. The server can use MySQL or a similar relational database as the storage unit to achieve the persistence of structured data; The server can use data processing libraries such as Pandas and NumPy to perform data preprocessing and feature extraction; The server can use TensorFlow or PyTorch as a deep learning framework for training and deploying generative artificial intelligence models and sentiment inference models. The server can use natural language processing libraries, such as NLTK or spaCy, to perform text segmentation, part-of-speech tagging, entity recognition, and dependency parsing. The server can communicate with external information sources via HTTP or HTTPS protocols, such as financial data service interfaces, public information data interfaces, and monitoring data interfaces.

[0349] The terminal can be a smart mobile terminal, an industrial control terminal, or a personal computing device. The terminal may be equipped with a display, touchscreen, microphone, camera, and network communication module (which may include a cellular communication module, a wireless LAN module, or a wired communication interface). The terminal may run a mobile operating system or a desktop operating system and have front-end applications or browser clients installed.

[0350] Users can interact with the server through the graphical user interface on the terminal, input attribute information, environmental information, and requirement information, and view the use cases or solutions output by the server, as well as the results related to the control of external devices.

[0351] II. Data Structures and Data Processing Methods Servers can design specific data structures for different types of data in storage units.

[0352] The server can maintain a relational table structure for user attribute information and environment information. Fields can include user ID, organization type, industry category, geographic information, network status indicators, terminal type, etc.

[0353] The server can maintain text fields and structured fields for the requirement information. Text fields are used to record natural language descriptions, while structured fields are used to record task types, target constraints, and priorities.

[0354] The server can create separate data tables for external information source data (such as financial data, publicly available case data, and sensor observation data). Fields can include data source identifiers, timestamps, data values, and quality markers.

[0355] The server can perform the following data processing on the above data through the data processing unit: The server can perform normalization on numerical data, such as using linear scaling or Z-score standardization, to map feature values ​​to a uniform numerical range, thereby improving the stability of neural network training. The server can perform one-hot encoding or embedding encoding on categorical data, converting the category identifiers into vector form so that they can be fed into the neural network; The server can perform word segmentation, stop word removal, stemming, and named entity recognition on natural language text, and convert the text into a fixed-length vector based on word embedding models (such as word vectors or context vectors). The server can build a time series structure for sensor observation data, sorting and segmenting the data according to timestamps for subsequent use in calculating control or setpoint values.

[0356] Through the aforementioned feature extraction process, the server transforms multi-source data from users, terminals, and external information sources into data to be processed. This data can be stored in memory using key-value mappings or multidimensional arrays for rapid access by generative AI models and rule engines.

[0357] III. Generative Artificial Intelligence Model Structure and Learning Methods The server can implement generative artificial intelligence models in the TensorFlow or PyTorch framework. These models can employ sequence-to-sequence neural networks based on the Transformer architecture.

[0358] The server can divide the model into two parts: an encoder and a decoder. The server can use multi-layer self-attention modules and feedforward networks in the encoder to encode the prompts and their additional features into a latent representation; The server can use multi-layer self-attention modules, cross-attention modules, and feedforward networks in the decoder to progressively generate sequence tags for output use cases or schemes based on the encoder's implicit representation.

[0359] The server can define the following learning elements for the model: The server can use the cross-entropy loss function as the main error function to measure the difference between the generated sequence and the manually labeled reference sequence; The server can use stochastic gradient descent or adaptive optimization algorithms to update network parameters; The server can use a batch training method, taking a large number of use case or solution texts from the historical sample library and their corresponding input features as training data; The server can use data augmentation techniques during training, such as synonym replacement, sentence order fine-tuning, and noise injection, to enhance the robustness of the model.

[0360] The server can load the trained model parameters into memory during deployment and perform forward propagation through an efficient inference engine, thereby reducing response latency. In some implementations, the server can leverage a graphics processing unit to accelerate matrix operations, thereby improving generation speed.

[0361] IV. Prompt Statement Generation and Rule Linkage The server can construct prompt text based on the data to be processed using the prompt generation unit. The prompt generation unit can perform the following detailed steps: The server can extract task type, key entities (such as industry, equipment type, scene location) and target constraints (such as latency limit, cost limit) from the data to be processed. The server can automatically generate prompts by filling these features into natural language templates based on a predefined template library.

[0362] For example, the server can generate the following prompt: "Based on the following data, please generate an integrated online and offline advertising strategy for the top 10 IT companies by sales revenue. The data includes: target audience profiles, advertising channels and performance metrics over the past 3 years." For example, the server can generate the following prompt: "The following is a summary of production line data for the past hour: Equipment A has stopped 3 times with a defect rate of 5%; Equipment B has a utilization rate of 95%; and there is overall production line congestion. Please generate 3 actionable optimization suggestions, including adjusting the cycle time, reallocating tasks, or adding buffer zones." For example, the server can generate the following prompt: "The target region is a rural area in a developing country with a moderate population density, unstable mobile network coverage, and an internet penetration rate of 30%. The main problems are a lack of educational resources and a shortage of doctors. Based on these conditions, please generate three feasible communication technology use cases and explain the implementation steps and expected results." After receiving use cases or solutions output by a generative artificial intelligence model, the server can parse the classification information and metadata into structured tags, including business category, scenario type, urgency level, etc. The server can then match these tags with a set of rules in the storage unit to determine which processing entity to classify the user information to and which control values ​​or settings need to be generated.

[0363] Through this linkage mechanism of "model output label - rule matching - processing subject selection", the server enables the generative artificial intelligence model to not be directly responsible for process control, but to work in conjunction with the rule set by outputting structured meta-information, thereby improving the interpretability and maintainability of the system.

[0364] V. Emotional Prediction and Prioritization Control The server can calculate user emotional information using speech, image, and character data through an emotion inference model. This model can employ a combination of convolutional neural networks and recurrent neural networks to process image frames and text sequences. The server can encode emotion categories (e.g., joy, anger, anxiety) and emotion intensity into numerical vectors.

[0365] The server can attach emotional information to the data to be processed and dynamically adjust the tone and output requirements when generating prompts. For example, when the emotion is "anger", the server can automatically increase the weight of response speed and reliability-related factors in the prompts.

[0366] The server can calculate priority scores based on sentiment information and the urgency of use cases or solutions within the priority control unit. Priority scores can be obtained through weighted summation, nonlinear functions, or ranking learning algorithms. The server can then reorder tasks in its internal queue based on these priority scores, ensuring that high-urgency and highly sentimental requests receive priority processing resources. This feature vector-based and numerically computed scheduling mechanism, compared to a simple first-come, first-served (FFS) model, can significantly reduce response latency for critical tasks and improve the utilization efficiency of computing and network resources.

[0367] VI. Calculation of control values ​​and setpoints and control of external devices Upon receiving a use case or solution, the server can calculate control or setpoint values ​​based on the technical steps and parameter suggestions within, combined with real-time observation data. The server can employ the following technical means: The server can use regression models or reinforcement learning policy networks to calculate continuous control parameters (such as equipment speed, buffer duration, and bandwidth allocation ratio) based on state characteristics (such as the current production line status or the current network load). The server can use constraint solvers or linear programming algorithms to find the optimal set of parameters for multi-device collaborative control, while adhering to physical and business constraints.

[0368] The terminal can receive control values ​​or setting values ​​sent by the server and convert them into specific control commands for actuators or network devices through the local driver. After executing the control commands, the terminal can continue to collect status data and feed it back to the server so that the server can evaluate the execution results and adjust the control strategy as necessary.

[0369] For example, when a user presents the following prompt: "Please analyze the current data of this automotive parts production line and provide specific suggestions for optimizing the cycle time of each robot and the inventory level of the buffer zone." The server can adjust the robot's working frequency and buffer capacity according to the scheme generated by the model, and calculate the new working cycle setting value for each robot; the terminal can write these setting values ​​into the controller, thereby changing the equipment operation mode on the actual production line.

[0370] VII. Feedback Information and Model Updates The server can receive user evaluation and feedback information from the terminal, such as user satisfaction ratings for a certain advertising strategy, changes in yield rates after production line adjustments, and latency statistics after network configuration adjustments. The server can then re-incorporate this information into the data to be processed and construct new training samples.

[0371] The server can periodically perform an incremental learning process: The server can extract recent execution records and feedback results from the storage unit and automatically construct input-output pairs for fine-tuning generative artificial intelligence models and sentiment inference models. The server can automatically or semi-automatically update the rule set based on feedback information, such as reducing the weight of rules with high error rates or enabling alternative rules.

[0372] Through this closed-loop update mechanism, the server can gradually adapt the generative artificial intelligence model and rule set to environmental changes without relying on frequent manual parameter tuning, thereby maintaining high accuracy and stability in long-term operation.

[0373] VIII. Technical Effects and Improvements in Computer Technology Through a series of specific technical processes, including multi-source data integration, feature extraction, automatic generation of prompts, linkage between generative artificial intelligence models and rule sets, emotion-driven scheduling, and external device control, the server not only automates business processes but also achieves improvements in the following computer technologies: By centrally managing multi-source heterogeneous data and unifying feature representation, the server reduces cross-system data conversion and redundant storage, improves data access locality and cache hit rate, thereby increasing the overall data processing speed. The server automatically generates prompts based on the current task and context, enabling generative AI models to receive structured and context-complete input, improving the relevance and accuracy of model output, reducing the number of multi-round interactive calls, and lowering network communication load. The server introduces a set of rules after the model output for secondary filtering and mapping, avoiding embedding complex process control logic inside the model, reducing model complexity, and facilitating model compression and inference acceleration. The server calculates priority scores by combining sentiment information and urgency information, and dynamically schedules the request queue. This can shorten the average response time of critical requests under limited computing resources and improve the technical performance of computing and network resource allocation strategies. The server uses a feedback-driven incremental learning mechanism to enable the generative artificial intelligence model and rule set to be updated adaptively, reducing the workload of manual rule maintenance and preventing performance degradation in the long run.

[0374] Because the various modules mentioned above are tightly coupled through specific data structures and algorithmic processes, the system can achieve improved accuracy, reduced latency, and increased resource utilization at the pure computational level, rather than simply automating manual operations. In this way, the server, within the computer, achieves technological improvements over traditional information processing methods through optimized feature encoding, model inference, and scheduling algorithms.

[0375] use Figure 14 The processing procedure is explained.

[0376] Step 1: The user enters information on the terminal. Users input attribute information, environmental information, and requirement information through a graphical user interface on the terminal. Input may include: industry category, organization size, location, network status description, device type, and target task description. The terminal then assembles this input into structured data and text data.

[0377] Input: Form data and natural language text entered by the user through the terminal.

[0378] Output: A request data packet generated within the terminal, containing key-value pair fields and raw text.

[0379] The terminal packages the data into a predefined format based on the input fields, such as a message object containing user identifiers, timestamps, attribute fields, and text fields, in preparation for subsequent transmission.

[0380] Step 2: The terminal sends the user request to the server. The terminal sends request packets to the server using the HTTPS protocol through a communication module (such as a cellular network or wireless LAN).

[0381] Input: The request data packet already packaged in the terminal.

[0382] Output: The network request message sent to the server.

[0383] Before sending, the terminal serializes and compresses the data packets as necessary, and adds authentication information to the message header to ensure that the server can identify the user and session.

[0384] Step 3: The server receives and parses the user request. The server receives HTTP requests from the terminal at the network interface and parses them at the application layer by the web framework.

[0385] Input: A network request message from the terminal.

[0386] Output: The original request object inside the server, which contains user attribute information, environment information, and requirement information.

[0387] The server uses a web framework to deserialize the message body into an internal data structure, converts structured fields into dictionaries or records, stores text fields as strings in memory, and assigns a unique request identifier.

[0388] Step 4: The server acquires and integrates data from external information sources. Based on the request content, the server obtains relevant structured and unstructured data from external information sources (such as business databases, public information databases, or monitoring systems) and integrates them with user information.

[0389] Input: Parsed user information and query conditions required to access external information sources.

[0390] Output: A comprehensive dataset containing user information and external data.

[0391] The server constructs a query based on industry fields, geographical location, and other conditions, calls an external service through a network interface, and retrieves the returned JSON or CSV data. Subsequently, the server uses a data processing library to convert the external data into a DataFrame, and associates them according to user identifiers or scenario identifiers to generate a unified data structure.

[0392] Step 5: The server performs normalization and feature extraction on multi-source data. The server performs numerical normalization, categorical encoding, and text vectorization on the comprehensive dataset to form the data to be processed.

[0393] Input: A comprehensive dataset containing user information and external data.

[0394] Output: Numericalized multidimensional feature vectors or feature sets.

[0395] The server performs linear scaling or normalization on numerical features and one-hot encoding or embedding index transformation on categorical features. For text fields, it performs word segmentation, stop word removal, and word embedding mapping, converting the text into vectors. The server concatenates all features into vectors or tensors in a predefined order and caches them as data to be processed.

[0396] Step 6: The server generates prompts for invoking the generative artificial intelligence model. The server automatically constructs prompts in natural language based on the data to be processed and the task conditions.

[0397] Input: The data to be processed and the condition information corresponding to the user's goal (such as task type and constraints).

[0398] Output: Prompt text used as input to the generative artificial intelligence model.

[0399] The server extracts key fields (such as industry, scenario, and target metrics) from the features, fills these fields into a template, for example: "Based on the following data, please generate use cases or solutions for the top-selling entities in a certain industry", and concatenates specific constraints to obtain a complete prompt statement.

[0400] Step 7: The server invokes a generative artificial intelligence model to generate use cases or solutions. The server sends the prompt to the generative AI model deployed on the server or in an external inference service and retrieves the output.

[0401] Input: The prompt text constructed by the server.

[0402] Output: Use case or solution text output by the generative artificial intelligence model.

[0403] The server encodes the prompt statement into a sequence of tags, performs forward inference using a deep learning framework, and generates an output sequence tag by tag. The server then performs post-processing on the generated results, such as removing redundant prefixes, standardizing the format, and truncating excessively long content.

[0404] Step 8: The server parses the model output and extracts classification information and metadata. The server performs semantic parsing on the use case or solution text, extracting meta-information such as category tags, urgency, and scope of application.

[0405] Input: Use case or solution text output by the generative artificial intelligence model.

[0406] Output: A set of structured classification information and meta-information.

[0407] The server uses natural language processing algorithms to extract keywords and match patterns in the text, mapping the text to predefined categories (such as "network optimization" and "production line control"), and identifying urgency, time requirements, etc. from the text to form structured tags that can be used for rule matching.

[0408] Step 9: The server classifies and determines the processing subject based on rule sets. The server uses a set of rules to map classification information and metadata to processing entities or control processes within the organization.

[0409] Input: Classification information and meta-information parsed from the model output, and a set of rules stored on the server.

[0410] Output: Determined processing entity identifier, target process identifier, and control strategy type.

[0411] The server matches tags according to rules, such as mapping "Category = Production Line Control and Urgency = High" to a specific control module; this process is completed through table lookup and logical judgment, generating the corresponding processing subject and control strategy instructions.

[0412] Step 10: The server acquires observation data and calculates control or setpoint values. The server calculates specific control parameters based on the use case or scheme and the current observation data.

[0413] Input: Use case or solution text, parsed control logic, and observation data sent by external devices or terminals.

[0414] Output: Control values ​​or setpoints used to control devices or configure systems.

[0415] Following the recommendations in the use case, the server extracts state variables from the observation data, uses algorithms (such as regression calculation, optimization solution, or policy network inference) to solve for the optimal control parameters, and organizes them into structured control instructions.

[0416] Step 11: The server sends control commands to the terminal. The server sends messages containing control or setting values ​​to the terminal via a communication network.

[0417] Input: Control values ​​or settings generated internally by the server and their target device identifier.

[0418] Output: Control command messages sent to the terminal.

[0419] The server encapsulates the control values ​​into a command format, adds information such as the device address and execution time, and sends it to the corresponding terminal through a secure communication protocol.

[0420] Step 12: The terminal receives control values ​​and drives external devices to perform operations. The terminal receives control commands from the server and converts control values ​​or set values ​​into specific operations for external devices.

[0421] Input: Control command messages sent by the server.

[0422] Output: Specific actions generated on external devices, such as parameter modification or state switching.

[0423] The terminal parses the message, extracts the target parameters, writes them to the local controller or device interface, and applies the new set values ​​to the actuator through the bus or control protocol, thereby changing the production line status, network configuration or other physical quantities in the real environment.

[0424] Step 13: Users can view test cases or schemes and execution results through the terminal. Users can view the use cases or solutions generated by the server on the terminal and observe the feedback information after the external device executes them.

[0425] Input: Use case or solution text obtained by the terminal from the server, and execution status data collected from external devices.

[0426] Output: A visual interface and human-computer understandable results presented to the user.

[0427] The terminal displays text content in the form of lists, paragraphs, or charts, and presents the execution status (such as success, failure, error value, performance indicators) in graphical or numerical form for user evaluation.

[0428] Step 14: Users input evaluations and feedback on the terminal. Users can input ratings, comments, or modification suggestions on the terminal based on the quality of the solution and its implementation effect.

[0429] Input: Subjective evaluations and supplementary text entered by the user in the terminal interface.

[0430] Output: A data packet containing evaluation and feedback information packaged by the terminal.

[0431] The terminal structures the ratings and comments, combining numerical ratings, tag selections, and text descriptions into a record, and adds a timestamp and request identifier so that the server can associate it with the corresponding use case or solution.

[0432] Step 15: The terminal sends evaluation and feedback information to the server. The terminal transmits user evaluations and execution feedback data to the server via the network.

[0433] Input: Evaluation and feedback information records stored internally in the terminal.

[0434] Output: The feedback message sent to the server.

[0435] The terminal uses the same session or authentication mechanism as the request to submit the feedback message to the specified interface of the server, ensuring that the feedback data can be accurately associated with the previous processing flow.

[0436] Step 16: The server incorporates the feedback data into the data to be processed and updates the model and rules. The server receives feedback information, writes it to the storage unit, and uses it to update the generative artificial intelligence model and rule set.

[0437] Input: Evaluation and feedback information sent by the terminal, as well as the original training data and rule set.

[0438] Output: Updated model parameters, adjusted rule entries, and new data records to be processed.

[0439] The server correlates the feedback with the original request and execution result, constructs new training samples, fine-tunes the generative artificial intelligence model using incremental training, and adjusts the rule weights based on error conditions or success rates, thereby improving generation quality and decision accuracy in subsequent request processing.

[0440] 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.

[0441] Data generation model 58 is a so-called generative AI (Artificial Intelligence). Examples of data generation models 58 include ChatGPT (registered trademark) (accessible via the internet (URL: https: / / openai.com / blog / chatgpt)). Data generation model 58 is obtained through deep learning on a neural network. Input to data generation model 58 are prompt words containing instructions, and inference data such as sound data representing sound, text data representing text, and image data representing images (e.g., still image data or animation data). Data generation model 58 infers from the input inference data based on the instructions represented by the prompt words and outputs the inference result in one or more data forms, such as sound data, text data, and image data. Data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or induction. The specific processing unit 290 performs the aforementioned specific processing while using data generation model 58. The data generation model 58 can also be a model finely tuned to output inference results from prompts that do not contain instructions. In this case, the data generation model 58 can output inference results based on prompts that do not contain instructions. The data processing apparatus 12, etc., includes various data generation models 58, including AI other than the generation AI. AI other than the generation AI can be, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and can perform various processes, but is not limited to this example. Furthermore, the AI ​​can also be an AI agent. Furthermore, when the processing of the above-mentioned parts is performed by AI, the processing can be performed partially or entirely by AI, but is not limited to this example. Furthermore, the processing performed by the AI ​​including the generation AI can be replaced by processing in the rule base, and the processing in the rule base can also be replaced by processing performed by the AI ​​including the generation AI.

[0442] 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.

[0443] 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.

[0444] 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.

[0445] Second Implementation Method Figure 3 An example of the configuration of the data processing system 210 according to the second embodiment is shown.

[0446] 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.

[0447] 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).

[0448] 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.

[0449] 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.

[0450] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and imaging elements such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, which captures images of the user 20's surroundings (e.g., the field of view defined by an angle equivalent to the field of vision of an average healthy person).

[0451] 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.

[0452] 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.

[0453] 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.

[0454] 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).

[0455] 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.

[0456] 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".

[0457] 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.

[0458] 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.

[0459] 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.

[0460] 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.

[0461] 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.

[0462] Data generation model 58 is a so-called generative AI (Artificial Intelligence). Examples of data generation models 58 include ChatGPT (registered trademark) (accessible via the internet (URL: https: / / openai.com / blog / chatgpt)). Data generation model 58 is obtained through deep learning on a neural network. Input to data generation model 58 are prompt words containing instructions, and inference data such as sound data representing sound, text data representing text, and image data representing images (e.g., still image data or animation data). Data generation model 58 infers from the input inference data based on the instructions represented by the prompt words and outputs the inference result in one or more data forms, such as sound data, text data, and image data. Data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or induction. The specific processing unit 290 performs the aforementioned specific processing while using data generation model 58. The data generation model 58 can also be a model finely tuned to output inference results from prompts that do not contain instructions. In this case, the data generation model 58 can output inference results based on prompts that do not contain instructions. The data processing apparatus 12, etc., includes various data generation models 58, including AI other than the generation AI. AI other than the generation AI can be, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and can perform various processes, but is not limited to this example. Furthermore, the AI ​​can also be an AI agent. Furthermore, when the processing of the above-mentioned parts is performed by AI, the processing can be performed partially or entirely by AI, but is not limited to this example. Furthermore, the processing performed by the AI ​​including the generation AI can be replaced by processing in the rule base, and the processing in the rule base can also be replaced by processing performed by the AI ​​including the generation AI.

[0463] 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.

[0464] 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.

[0465] 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.

[0466] Third Implementation Method Figure 5 An example of the configuration of the data processing system 310 according to the third embodiment is shown.

[0467] 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.

[0468] 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).

[0469] 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.

[0470] 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.

[0471] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and imaging elements such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, which captures images of the user 20's surroundings (e.g., the field of view defined by an angle equivalent to the field of vision of an average healthy person).

[0472] 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.

[0473] 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.

[0474] 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.

[0475] 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.

[0476] 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.

[0477] 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".

[0478] 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.

[0479] 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.

[0480] 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.

[0481] 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.

[0482] 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.

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

[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 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.

[0485] For example, the collection unit is implemented by the control unit 46A of the head-mounted terminal 314 or the specific processing unit 290 of the data processing device 12. For example, the acquisition unit uses the camera 42 or communication I / F 44 of the head-mounted terminal 314 to acquire step data, which is then processed by the specific processing unit 290 of the data processing device 12. For example, the analysis unit is implemented by the specific processing unit 290 of the data processing device 12 to analyze the data from the collection unit and the acquisition unit. For example, the generation unit is implemented by the specific processing unit 290 of the data processing device 12 to generate a menu using a generation AI. For example, the serving unit is implemented by the speaker 240 and display 343 of the head-mounted terminal 314 or the specific processing unit 290 of the data processing device 12 to provide the generated menu to the user. The correspondence between each unit and the device or control unit is not limited to the above examples and various changes can be made.

[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 head-mounted terminal 314.

[0487] Fourth Implementation Method Figure 7 An example of the configuration of the data processing system 410 according to the fourth embodiment is shown.

[0488] 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.

[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] 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.

[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 photograph the area around robot 414 (e.g., the field of view defined by a perspective equivalent to the field of vision of an average healthy person).

[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] 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.

[0495] 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.

[0496] 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.

[0497] 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.

[0498] 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.

[0499] 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".

[0500] 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.

[0501] 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.

[0502] 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.

[0503] 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.

[0504] 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.

[0505] Data generation model 58 is a so-called generative AI (Artificial Intelligence). Examples of data generation models 58 include ChatGPT (registered trademark) (accessible via the internet (URL: https: / / openai.com / blog / chatgpt)). Data generation model 58 is obtained through deep learning on a neural network. Input to data generation model 58 are prompt words containing instructions, and inference data such as sound data representing sound, text data representing text, and image data representing images (e.g., still image data or animation data). Data generation model 58 infers from the input inference data based on the instructions represented by the prompt words and outputs the inference result in one or more data forms, such as sound data, text data, and image data. Data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or induction. The specific processing unit 290 performs the aforementioned specific processing while using data generation model 58. The data generation model 58 can also be a model finely tuned to output inference results from prompts that do not contain instructions. In this case, the data generation model 58 can output inference results based on prompts that do not contain instructions. The data processing apparatus 12, etc., includes various data generation models 58, including AI other than the generation AI. AI other than the generation AI can be, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and can perform various processes, but is not limited to this example. Furthermore, the AI ​​can also be an AI agent. Furthermore, when the processing of the above-mentioned parts is performed by AI, the processing can be performed partially or entirely by AI, but is not limited to this example. Furthermore, the processing performed by the AI ​​including the generation AI can be replaced by processing in the rule base, and the processing in the rule base can also be replaced by processing performed by the AI ​​including the generation AI.

[0506] 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.

[0507] 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.

[0508] 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.

[0509] Furthermore, the emotion-specific model 59, acting as an emotion engine, can determine a user's emotion based on a specific mapping. Specifically, the emotion-specific model 59 can determine a user's emotion based on an emotion graph that serves as a specific mapping (see [reference]). Figure 9 The emotion-specific model 59 can also determine the robot's emotion, and the specific processing unit 290 performs specific processing based on the robot's emotions.

[0510] 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.

[0511] 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.

[0512] 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).

[0513] 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.

[0514] 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."

[0515] 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.

[0516] 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).

[0517] 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.

[0518] 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.

[0519] 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.

[0520] 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.

[0521] 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.

[0522] 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.

[0523] 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.

[0524] 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.

[0525] 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.

[0526] 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.

[0527] In addition, the following notes are provided in response to the above explanation.

[0528] Example 1 (Note 1) An information processing system, characterized in that it comprises: A means for receiving identification information from a communication device indicating the location of a publicly available object; A device for communicating with an external information processing device based on the identification information and automatically acquiring standardized descriptive information from multiple information sources; An apparatus for parsing the descriptive structure of the descriptive information, extracting textual information from unstructured information, and generating a sequence of information elements; A device for preprocessing the information element sequence for input into a generative artificial intelligence model, and for converting the information element sequence into feature information through lexicalization and encoding. A means for performing inference processing of a generative artificial intelligence model using the feature information to generate evaluation information representing the relevance of wireless communication technology to instances and organizational attributes; A means for identifying organizations that are active in the use of wireless communication technologies based on the evaluation information, and for generating organization index information representing the correspondence between the organization and overseas cases. An apparatus for selecting a sentence structure pattern for automatically generating prompts for input to a generative artificial intelligence model based on the organization index information and the evaluation information, and for generating query text that includes at least one of the following: technology type, industry type, and region type. A device for sending the query text to the communication device and generating control information for inputting the query text into a generative artificial intelligence model based on input operations from the communication device; An apparatus for storing the response information obtained from the generative artificial intelligence model in correspondence with the organization index information and the evaluation information, and for accumulating knowledge information that can be reused in subsequent query text generation.

[0529] (Note 2) The information processing system according to Appendix 1 is characterized in that, The communication device is configured to: display candidates of the query text in a list on a display device, accept selection or editing operations from the user, update the query text according to the operation result, and perform control when the updated query text is sent to the generative artificial intelligence model.

[0530] (Note 3) The information processing system according to Appendix 1 is characterized in that, The processor is configured to: calculate an index value representing the utilization level of wireless communication technology application cases by industry type or technology type based on the response information output from the generative artificial intelligence model and the evaluation information, and dynamically change the priority of the technology type or industry type contained in the query text based on the index value to generate the prompt statement.

[0531] Application Example 1 (Note 1) An information processing system, characterized in that it comprises: A unit executed by a processing unit in an information processing device, which acquires information from a user and receives the information via an input / output device; A unit that generates a retrieval string for obtaining public information through a communication medium based on the user information and pre-set conditions; The unit that retrieves public information from external information sources based on the search string and extracts text information from the public information according to predetermined information extraction conditions; Based on the user information, the extracted text information, and the pre-set business conditions, a unit is generated to produce prompt statements for the generative artificial intelligence model to perform parsing processing or case generation processing. The prompt statement and the extracted text information are input into the generative artificial intelligence model, and the instance information and organizational information units generated by the generative artificial intelligence model are obtained. The unit that calculates performance indicators based on the organizational information and sorts the organizations based on the performance indicators to generate an organizational list information. Based on the case information, the organization list information, and the pre-defined business model information, a unit of application case information that can be applied to user business is generated; The application example information is sent to a terminal device that includes a portable display device or a wearable display device, and the application example information is converted into display data that can be displayed in real time in the terminal device. The unit that acquires the user's voice input through the terminal device and generates additional search strings or additional prompt statements based on the voice input, thereby performing cyclic control on the parsing processing or case generation processing of the generative artificial intelligence model. Based on the application case information, a unit is generated to determine the classification results of the information obtained from the user into different business categories or service categories according to a predetermined classification rule; Based on the determination result, the unit that allocates the information obtained from the user according to the business classification or service classification.

[0532] (Note 2) The information processing system according to Appendix 1 is characterized in that, The information processing device includes: a unit that evaluates the emotional state by taking voice data, text data and image data obtained from the user as input, generates an analysis result corresponding to the emotional state, and controls and changes the content of the prompt statement or the prompt form of the example information based on the analysis result.

[0533] (Note 3) The information processing system according to Appendix 1 is characterized in that, The information processing device includes: a control unit that assigns priority indicators to the information obtained from the user and the application case information according to the emotional state analysis results described in Appendix 2 and the case information generated by the generative artificial intelligence model, and determines the processing order of the information and the order of the information output to the terminal device according to the priority indicators.

[0534] Example 2 (Note 1) An information processing system, characterized in that it comprises: A device for receiving natural language instructions from a user via an input / output device; An apparatus for sending the instruction information as a prompt statement to a generative artificial intelligence model, and instructing the generative artificial intelligence model to generate structured conditional information including business domain, time range, and evaluation items based on the content of the instruction information; An apparatus for generating acquisition conditions for multiple target information stored in an information storage device based on the structured condition information, and for extracting the target information by performing information management functions through an information processing device; A device for performing information processing programs on extracted target information, including handling missing values, removing outliers, and unifying data types, to generate regularized information. An apparatus for performing numerical attribute-based sorting and extraction quantity restriction processing on the regularized information to generate a set of target elements assigned an order as specific information. A control processing device for repeatedly executing the extraction, normalization, and sorting processes, which prompts the user with specific information through the input / output device and dynamically updates the structured condition information and the acquisition conditions based on additional prompts for the generative artificial intelligence model.

[0535] (Note 2) The information processing system according to Appendix 1 is characterized in that, The target information specified in the structured condition information includes financial information as business information. The normalization process includes converting the numerical items of the financial information into a unified unit of measurement and converting different currency units. The set of target elements assigned a sequence includes a set of organizations configured at the top level based on sales amount.

[0536] (Note 3) The information processing system according to Appendix 1 is characterized in that, The control processing device is configured to: send a prompt statement with user input to the generative artificial intelligence model for condition parsing, and automatically generate filtering conditions, sorting conditions, and output format conditions for the regularized information based on the condition parsing results; regenerate the target element set with assigned order according to the automatically generated conditions, and convert the obtained results into data information for reporting and output them.

[0537] Application Example 2 (Note 1) An information processing system, characterized in that it comprises: A receiving unit for receiving attribute information, environmental information, and demand information from a user via an input / output device; A data processing unit is used to store received information, as well as structured and unstructured data obtained from external information sources, into a storage unit within an information processing device, and to integrate, normalize, and extract features from the data to generate data to be processed. A prompt generation unit is used to automatically generate prompt statements for input into a generative artificial intelligence model based on the data to be processed and condition information corresponding to the user's purpose, and to issue an instruction to the generative artificial intelligence model to input the prompt statements to generate use cases or solutions. A classification unit for assigning and classifying the user's information to the processing entity within the organization based on the classification information and meta-information contained in the use cases or schemes output from the generative artificial intelligence model, and the rule set stored in the storage unit. Used to calculate control values ​​or setpoints based on observation data obtained from external devices, and to send the control values ​​or setpoints to the control unit of the terminal device via a communication network, based on the use case or scheme. An update unit for receiving user evaluation information or feedback information from the terminal device and importing it again as the data to be processed, in order to update the generative artificial intelligence model or the rule set.

[0538] (Note 2) The information processing system according to Appendix 1 is characterized in that the system further includes: An emotion processing unit is used to acquire voice data, image data, and character data from a voice acquisition device, an image acquisition device, and a character input device; execute an emotion estimation model for estimating the user's emotional state; thereby generating emotion information representing the emotional state; attach the emotion information to the data to be processed; and dynamically change the content of the prompt statement or the generation conditions of the use case or scheme based on the emotion information.

[0539] (Note 3) The information processing system according to Appendix 1 is characterized in that the system further includes: A priority control unit for calculating a priority score based on the emotional information and the urgency information contained in the use case or scheme, and for determining the processing order of the user information and the application order of the control value or setting value according to the priority score.

Claims

1. An information processing system, characterized in that, include: processor; The processor is configured to: receive information through an interface for receiving information from a user; generate prompts to instruct a generative artificial intelligence model to parse the received information and generate relevant use cases; and, based on the generated use cases, apply pre-defined rules to sort the information to the corresponding departments or services.

2. The information processing system according to claim 1, characterized in that, The processor is also configured to: analyze the user's emotional state using voice data, text data, and facial expression data as input, assess the user's emotions, and generate emotion analysis results.

3. The information processing system according to claim 1, characterized in that, The processor is also configured to: adjust the priority of information based on the sentiment analysis results, calculate a priority score, and determine the processing order of information according to the priority score.

Images