Intelligent calling data method and device based on semantic understanding

Abstract

This application provides a semantic understanding-based intelligent data retrieval method and apparatus, comprising: responding to a current user instruction input by a user, generating structured text information based on the current user instruction; determining semantic information and target resources based on the structured text information, and generating task analysis results and task classification tags through a preset supervisory agent based on the semantic information and target resources, wherein the semantic information includes user intent, instruction entity, and contextual relationship; selecting the corresponding processing module in the business system through the supervisory agent based on the task classification tags, and transmitting the task analysis results to the corresponding processing module; generating structured response instruction information through an instruction control sub-agent based on the task analysis results, and retrieving target data based on the response instruction information, thereby achieving accurate identification of user instructions, improving the response speed of user instruction execution, and constructing an efficient, secure, and convenient digital human intelligent data retrieval system.

Description

Technical Field

[0001] This invention relates to the fields of digital human interaction, natural language processing, and cross-platform invocation technology, specifically to an intelligent data retrieval method and apparatus based on semantic understanding. Background Technology

[0002] With the deep integration of digital human technology and industry information platforms, various professional fields are gradually introducing digital humans as an interaction portal to improve user convenience and experience. While digital humans, as natural interaction carriers, possess anthropomorphic communication capabilities, most current digital human systems have significant limitations in terms of function invocation in professional application scenarios such as slope monitoring.

[0003] The digital human system only supports basic interaction and lacks deep semantic parsing and automated execution capabilities; intent recognition relies on simple rules or models, which cannot handle ambiguous instructions and technical terms, resulting in low accuracy; the resource library is statically mapped and cannot dynamically adapt to system updates; and the process is semi-automated, requiring users to manually confirm and operate, failing to achieve closed-loop automation of "user input instruction - system automatic execution - result presentation," which severely limits the advantages and effectiveness of the digital human system in efficient interaction. Summary of the Invention

[0004] In view of the aforementioned problems, this application is proposed to provide a semantic understanding-based intelligent data retrieval method and apparatus for overcoming or at least partially solving the aforementioned problems, comprising: A semantic understanding-based intelligent data retrieval method, wherein the method utilizes digital human technology and a large language model combined with a multi-agent collaborative architecture to perform semantic understanding of user commands, the method comprising: In response to the current user instruction input by the user, structured text information is generated based on the current user instruction; Based on the structured text information, semantic information and target resources are determined, and through a preset supervisory agent, task analysis results and task classification labels are generated based on the semantic information and target resources. The semantic information includes user intent, instruction entities, and contextual relationships. Based on the task classification tags, the supervisory agent selects the corresponding processing module in the business system and transmits the task analysis results to the corresponding processing module. Based on the task analysis results, the sub-agent is controlled by preset instructions to generate structured response instruction information, and the target data is invoked based on the response instruction information.

[0005] Specifically, the step of generating structured text information based on the current user instruction in response to user input includes: The current user instructions are received through the front-end interface, wherein the current user instructions include user voice instructions and text input instructions; The current user command is preprocessed according to preset command format rules; The preprocessed current user instruction is subjected to format verification and security checks, and the corresponding structured text information is generated based on the current user instruction that conforms to the processing specifications.

[0006] Specifically, the steps of determining semantic information and target resources based on the structured text information, and generating task analysis results and task classification labels based on the semantic information and target resources through a preset supervisory agent, wherein the task analysis results include user intent, instruction entities, and contextual relationships, include: Based on the structured text information, semantic parsing is performed using the preset large language model to generate semantic information including the user intent, the instruction entity, and the context relationship; Based on the structured text information and the preset target resource library, a comprehensive constraint matching score is calculated using a multi-constraint weight allocation formula. Based on the comprehensive constraint matching score, the target resources are selected and extracted from the target resource library. Based on the semantic information and the target resources, the supervisory agent generates the task analysis results and the task classification labels.

[0007] Specifically, the step of calculating a comprehensive constraint matching score based on the structured text information and a preset target resource library using a multi-constraint weight allocation formula, and then filtering and extracting the target resources from the target resource library based on the comprehensive constraint matching score, includes: Based on the structured text information and the preset target resource library, the time constraint matching degree, location constraint matching degree and data type constraint matching degree are obtained respectively; Based on the time constraint matching degree, the location constraint matching degree, and the data type constraint matching degree, the comprehensive constraint matching score is calculated using the multi-constraint weight allocation formula; Based on the comprehensive constraint matching score, the target resources are selected and extracted from the target resource library.

[0008] Specifically, the step of generating the task analysis result and the task classification label by the supervisory agent based on the semantic information and the target resource includes: Based on the semantic information and the target resources, the supervisory agent performs deep analysis to generate the task analysis results; Based on the task analysis results and the predefined task classification system, task types are identified, and task classification labels are generated.

[0009] Specifically, the step of selecting the corresponding processing module in the business system based on the task classification label by the supervisory agent, and transmitting the task analysis results to the corresponding processing module, includes: Based on the task classification tags, the supervisory agent identifies and selects the corresponding processing module from the business system. Based on the task classification tags and the processing module, a routing strategy is generated through a preset intelligent routing model; According to the routing strategy, the task analysis results are transmitted to the processing module through the supervisory agent; Specifically, the step of generating structured response instruction information by controlling the sub-agent through preset instructions based on the task analysis results, and calling target data based on the response instruction information, includes: The sub-agent receives and parses the task analysis results through the instruction control, and generates the response instruction information. Based on the response instruction information, the target data is retrieved from the business system.

[0010] Once the target data has been retrieved, the instruction execution results and performance metrics are recorded, and an execution log is generated.

[0011] A semantic understanding-based intelligent data retrieval device includes: The user instruction preprocessing module, in response to the current user instruction input by the user, generates structured text information based on the current user instruction; The semantic parsing and task generation module determines semantic information and target resources based on the structured text information, and generates task analysis results and task classification tags based on the semantic information and target resources through a preset supervisory agent. The task analysis results include user intent, instruction entities, and contextual relationships. The task allocation and transmission module selects the corresponding processing module through the supervisory agent based on the task classification label, and transmits the task analysis results to the corresponding processing module. The instruction execution module, based on the task analysis results, controls the sub-agent to generate structured response instruction information through preset instructions, and calls the target data based on the response instruction information.

[0012] A computer electronic device includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein when the computer program is executed by the processor, it implements the steps of the semantic understanding-based intelligent data retrieval method as described above.

[0013] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the semantic understanding-based intelligent data retrieval method as described above.

[0014] This application has the following advantages: In the embodiments of this application, addressing the problems of existing digital human systems that only support basic interactions, lack deep semantic parsing and automated execution capabilities, cannot handle ambiguous instructions and technical terms, have static mapping of resource libraries leading to an inability to dynamically adapt to system updates, and require user intervention for manual confirmation and operation, this application provides a solution through a large language model and a multi-constraint weight allocation mechanism, combined with a multi-agent collaborative structure. Specifically, it involves: "Responding to the current user instruction input by the user, generating structured text information based on the current user instruction; determining semantic information and target resources based on the structured text information, and generating task analysis results and task classification tags through a preset supervisor agent based on the semantic information and target resources, wherein the task analysis results include user intent, instruction entities, and contextual relationships; selecting the corresponding processing module through the supervisor agent based on the task classification tags, and transmitting the task analysis results to the corresponding processing module; generating structured response instruction information through a preset instruction control sub-agent based on the task analysis results, and calling target data based on the response instruction information." The user instructions are semantically parsed using the pre-defined large language model. Combined with the pre-defined target database and multi-constraint weight allocation formula, the current user instructions are fully extracted and the user intent, task entity, and contextual relationship are accurately identified. By constructing a multi-agent collaborative architecture, the instruction control sub-agent is established in each processing module, and a master agent is set up to accurately classify and allocate tasks, forming a closed-loop process of "instruction input - automatic execution - result presentation". This completely eliminates the need for manual confirmation by the user, significantly improving operational efficiency compared to existing semi-automatic solutions. It solves the technical problem that current digital human systems can only achieve simple basic interactions and cannot fully extract, accurately identify, and automatically execute fuzzy instructions and professional terms based on the user's natural language. It establishes an automated mapping relationship between semantic parsing results and business system operations, realizing deep integration of digital human interaction scenarios and business operations. This achieves accurate identification of user instructions, improves the response speed of user instruction execution, and provides a user experience that better meets the expectations of human-like interaction. It satisfies the need for efficient interaction between digital humans and users, and constructs an efficient, secure, and convenient digital human-driven intelligent call system for web pages and data charts. It is applicable to the field of slope monitoring and can be directly extended to similar industrial monitoring scenarios such as tailings dam monitoring, bridge health monitoring, and tunnel construction monitoring. Only the domain-specific terminology dictionary and resource mapping library need to be replaced to achieve rapid adaptation. Attached Figure Description

[0015] To more clearly illustrate the technical solution of this application, the drawings used in the description of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the steps of an intelligent data retrieval method based on semantic understanding, provided in one embodiment of this application. Figure 2 This is a structural block diagram of an intelligent data retrieval device based on semantic understanding, provided in one embodiment of this application; Figure 3 This is a flowchart illustrating the workflow of an intelligent data retrieval device based on semantic understanding, provided in one embodiment of this application. Figure 4 This is a schematic diagram of the structure of a computer electronic device provided in an embodiment of the present invention; 1. A computer electronic device; 2. An external device; 3. A processing unit; 4. A bus; 5. A network adapter; 6. An I / O interface; 7. A display; 8. Memory; 9. Random access memory; 10. A cache memory; 11. A storage system; 12. A program / utility; 13. A program module. Detailed Implementation

[0017] To make the objectives, features, and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0018] The inventors, through analysis of existing technologies, discovered that in professional application scenarios such as slope monitoring, most current digital human systems have significant limitations at the function invocation level. Existing technologies suffer from low semantic parsing accuracy. They employ general keyword matching or basic classification models, failing to adapt to the specialized terminology of slope monitoring and lacking the ability to parse multi-constraint commands. This results in the system's inability to accurately identify certain specialized terms in the slope monitoring field, its difficulty in fully extracting multi-dimensional constraint information such as time, location, and data type, and its direct return of invalid results when faced with ambiguous commands. Ultimately, this leads to misinterpretation of command intent, incomplete extraction of key information, and an inability to meet the parsing requirements of specialized scenarios.

[0019] Traditional intelligent navigation and data query systems lack an automated mapping relationship between semantic parsing results and business system operations, and there is a lack of standardized calling patterns between the digital human system and the slope monitoring website. This results in the system only returning webpage links or report paths, requiring users to manually complete the final operation, leading to cumbersome operation processes, a disconnect between interaction and execution, and failing to leverage the core advantage of convenient interaction of digital humans.

[0020] Existing technologies primarily rely on pop-ups and links for interaction, failing to support multimodal command processing and human-like feedback. Furthermore, execution results cannot be embedded into the digital human's interactive interface. This results in incompatibility between the system and core scenarios such as digital human voice interaction and visualization, leading to a poor user experience and hindering deep integration between digital humans and business platforms.

[0021] It should be noted that, in any embodiment of the present invention, in response to the core requirement of integrating digital humans with industry platforms in the field of slope monitoring, the user commands are semantically understood and automatically executed based on digital human technology and a large language model combined with the multi-agent collaborative architecture.

[0022] Reference Figure 1 This illustrates an embodiment of an intelligent data retrieval method based on semantic understanding provided in this application: S110. In response to the current user instruction input by the user, generate structured text information based on the current user instruction; S120. Based on the structured text information, determine semantic information and target resources, and through a preset supervisory agent, generate task analysis results and task classification labels based on the semantic information and target resources. The task analysis results include user intent, instruction entities, and contextual relationships. S130. Based on the task classification label, the supervisory agent selects the corresponding processing module and transmits the task analysis result to the corresponding processing module. S140. Based on the task analysis results, the sub-agent is controlled by preset instructions to generate structured response instruction information, and the target data is called according to the response instruction information. In the embodiments of this application, the existing digital human system has significant limitations in terms of function call technology, which cannot fully extract user instructions and accurately parse semantics in professional scenarios, and cannot construct a fully closed-loop process in which the digital human automatically executes control operations based on user instructions. The technical solution of this application performs semantic parsing of the user instructions through a preset large language model, combined with a preset target database and a multi-constraint weight allocation formula, to fully extract the current user instructions and accurately identify the user intent, task entity and contextual relationship. By constructing a multi-agent collaborative architecture, the instruction control sub-agent is established in each processing module, and a master agent is set up to accurately classify and allocate tasks, forming a closed-loop process of "instruction input - automatic execution - result presentation". This completely eliminates the need for manual confirmation by the user, significantly improving operational efficiency compared to existing semi-automatic solutions. It solves the technical problem that current digital human systems can only achieve simple basic interactions and cannot fully extract, accurately identify, and automatically execute fuzzy instructions and professional terms based on the user's natural language. It establishes an automated mapping relationship between semantic parsing results and business system operations, realizing deep integration of digital human interaction scenarios and business operations. This achieves accurate identification of user instructions, improves the response speed of user instruction execution, and provides a user experience that better meets the expectations of human-like interaction. It satisfies the need for efficient interaction between digital humans and users, and constructs an efficient, secure, and convenient digital human-driven intelligent call system for web pages and data charts. It is applicable to the field of slope monitoring and can be directly extended to similar industrial monitoring scenarios such as tailings dam monitoring, bridge health monitoring, and tunnel construction monitoring. Only the domain-specific terminology dictionary and resource mapping library need to be replaced to achieve rapid adaptation.

[0023] The following will further describe an intelligent data retrieval method based on semantic understanding in this exemplary embodiment.

[0024] In one embodiment of the present invention, the specific process of "generating structured text information in response to the current user instruction input by the user" in step S110 can be further described in conjunction with the following description.

[0025] As described in the following steps The current user instructions are received through the front-end interface, and the current user instructions include user voice instructions and text input instructions. The current user command is preprocessed according to preset command format rules; After performing format validation and security checks on the preprocessed current user command, the corresponding structured text information is generated based on the current user command that conforms to the processing specifications. It should be noted that the user instructions include user voice instructions and text input instructions. The front-end interface establishes a unified instruction receiving interface to receive the user instructions and record metadata such as instruction source, timestamp, and user identifier.

[0026] In one specific implementation, when a user's voice command is received, the start and end points of the voice are accurately identified through voice activity detection technology to reduce environmental noise interference. The voice is then converted into text using a preset voice recognition system that supports multilingual and dialect recognition. The recognition results are corrected and post-processed to generate structured text information for semantic analysis. The structured text information retains auxiliary information such as intonation and pauses.

[0027] By integrating multimodal interaction and visualization, it supports voice and text multimodal command input, optimizes the collaborative process of ASR (Automatic Speech Recognition) transcription and semantic parsing, and provides human-like feedback such as digital human voice broadcasting and action demonstrations, combined with the visualization of execution results, deeply adapting to digital human interaction scenarios.

[0028] In one embodiment of the present invention, the specific process of step S120, which involves "determining semantic information and target resources based on the structured text information, and generating task analysis results and task classification labels based on the semantic information and target resources through a preset supervisory agent, wherein the task analysis results include user intent, instruction entities, and contextual relationships," can be further explained in conjunction with the following description.

[0029] As described in the following steps Based on the structured text information, semantic parsing is performed using the preset large language model to generate semantic information including the user intent, the instruction entity, and the context relationship; Based on the structured text information and the preset target resource library, a comprehensive constraint matching score is calculated using a multi-constraint weight allocation formula, and the target resources are extracted from the target resource library based on the comprehensive constraint matching score. Based on the semantic information and the target resources, the supervisory agent generates the task analysis results and the task classification labels.

[0030] It should be noted that, based on the user intent, the instruction entity, and the contextual relationship, combined with the target resources obtained through the multi-constraint weight allocation formula, the comprehensive evaluation of the three types of constraints—time, location, and data type—is the key to achieving complete extraction and accurate parsing of multi-constraint instructions and accurate matching of multi-dimensional instructions.

[0031] As an example, this method utilizes a technical terminology keyword similarity formula based on cosine similarity to achieve accurate matching within a technical terminology database for slope monitoring. The technical terminology keyword similarity formula is shown below:

[0032] in, User command keywords Keywords in the professional terminology database Similarity (0≤ ≤1); for In the dimension value of the word vector space, for Dimensional value in the word vector space; Add a domain correction factor (1.2, to strengthen the weight of matching technical terms). The dimension is the word vector.

[0033] This formula compares and verifies the similarity between the technical terms and keywords in user commands and the data in a pre-set database of technical terms for slope monitoring, thus solving the problem of ambiguous technical terminology recognition and improving the matching accuracy between commands and target web pages or charts.

[0034] As an example, an "enhanced single model + rule engine" architecture is adopted. By embedding a rule base (including a professional terminology dictionary and constraint extraction rules) in the slope monitoring domain into the large language model, and by pre-defining professional terminology mapping rules (such as "microseismic frequency" → corresponding data field) and multi-constraint extraction templates, the instruction parsing logic is optimized using Prompt Engineering. After the large language model LLM generates the parsing results, the rule engine performs secondary verification and completion. Without the need for an independent intelligent agent module, semantic parsing with the same accuracy is achieved, achieving the same parsing effect as the multi-agent architecture.

[0035] In one embodiment of the present invention, the specific process of "generating a comprehensive constraint matching score by means of a multi-constraint weight allocation formula based on the structured text information and a preset target resource library, and filtering and extracting the target resources from the target resource library based on the comprehensive constraint matching score" can be further described in conjunction with the following description.

[0036] As described in the following steps Based on the structured text information and the preset target resource library, the time constraint matching degree, location constraint matching degree and data type constraint matching degree are obtained respectively; Based on the time constraint matching degree, the location constraint matching degree, and the data type constraint matching degree, the comprehensive constraint matching score is calculated using the multi-constraint weight allocation formula; Based on the comprehensive constraint matching score, the target resources are selected and extracted from the target resource library.

[0037] It should be noted that the multi-constraint weight allocation formula is used to calculate the comprehensive weight of the three types of constraints: time, location, and data type. The multi-constraint weight allocation formula is shown below:

[0038] in, For the comprehensive constraint matching score (0≤ ≤1); (0.3) represents the time weighting coefficient. (0.4) represents the point weight coefficient. (0.3) represents the data type weighting coefficient; For time-constrained matching degree; For point-to-point constraint matching degree; Constrain the matching degree for data types.

[0039] The time weight coefficient, the location weight coefficient, and the data type weight coefficient are set based on the requirements of the slope monitoring scenario. The time constraint matching degree, the location constraint matching degree, and the data type constraint matching degree are determined based on the fit between the structured text information and the resource data in the target resource library.

[0040] In one specific implementation, the structured text information is compared and analyzed with the data in the target resource library to obtain the constraint matching degree value. Combined with the constraint weight coefficient preset according to the actual use scenario requirements, the comprehensive constraint matching score of each piece of data in the target resource library is obtained. Based on the comprehensive constraint matching score, the target resource that best matches the structured text information in terms of data, location, and data type is selected from the target resource library. This achieves accurate matching of multi-dimensional instructions and complete parsing of multi-constraint instructions, accurately identifying user intent, corresponding entity of instruction, and contextual relationship.

[0041] As an example, the time constraint matching degree can be, for example, the degree of fit between "yesterday" and the system time; the data type constraint matching degree can be, for example, the degree of fit between "3D scatter plot" and the chart type library.

[0042] In one embodiment of the present invention, the specific process of "generating the task analysis result and the task classification label by the supervisory agent based on the semantic information and the target resource" can be further described in conjunction with the following description.

[0043] As described in the following steps Based on the semantic information and the target resources, the supervisory agent performs deep analysis to generate the task analysis results; Based on the task analysis results and the preset task classification system, task types are identified, and task classification labels are generated.

[0044] It should be noted that the predefined task classification system includes task types such as command control, slope data query, and general question and answer, and different task types have corresponding dedicated processing modules.

[0045] In one specific implementation, the supervisory agent performs in-depth analysis of the semantic information and the target resources, accurately classifies the tasks according to the task classification system, determines the task type to which the current user instruction belongs, and generates the corresponding task classification label based on the task type.

[0046] In one embodiment of the present invention, the specific process of step S130, "selecting the corresponding processing module through the supervisory agent based on the task classification label and transmitting the task analysis result to the corresponding processing module," can be further explained in conjunction with the following description.

[0047] As described in the following steps Based on the task classification tags, the supervisory agent identifies and selects the corresponding processing module from the business system; Based on the task classification tags and the processing module, a routing strategy is generated through a preset intelligent routing model; According to the routing strategy, the task analysis results are transmitted to the processing module through the supervisory agent.

[0048] It should be noted that the processing modules are independent units in the business system responsible for executing specific task functions. The intelligent routing model dynamically determines the optimal path based on the task classification labels, system load, and module execution capacity.

[0049] As an example, after the routing strategy is generated, the task analysis results are encapsulated and verified through a relay function, security tokens and integrity checks are added, and a real-time bidirectional communication channel is established using WebSocket to ensure low-latency transmission and state synchronization between the supervisor agent and the instruction control sub-agent. Secure cross-domain message passing is achieved through a cross-domain communication mechanism between front-end pages, and message retry, acknowledgment, and error recovery mechanisms are supported to ensure the reliability and security of instruction transmission.

[0050] In one specific implementation, the supervisory agent selects the corresponding processing module from the business system based on the task classification label. The intelligent routing model generates the routing strategy based on the task classification label as the starting point and the corresponding processing module as the ending point, and routes the task analysis result to the processing module. The supervisory agent establishes task priority and execution queue based on the task analysis result and the task classification label.

[0051] As an example, an "MQTT message queue + API gateway" approach is used to replace WebSocket and iframe for cross-domain data transmission. The API gateway centrally manages access permissions for the slope monitoring system, enables asynchronous communication between the digital human system and the business system via the MQTT protocol, and routes encapsulated commands to the target module through the gateway. The MQTT protocol ensures low latency and resumable interrupted transmissions, while the API gateway handles security verification and protocol conversion. Cross-domain data transmission can be completed without relying on iframes, and automatic command execution and result feedback are also achieved.

[0052] In one embodiment of the present invention, the specific process of step S140, which involves "generating structured response instruction information by controlling the sub-agent through preset instructions based on the task analysis results, and calling target data based on the response instruction information," can be further explained in conjunction with the following description.

[0053] As described in the following steps The sub-agent receives and parses the task analysis results through the instruction control, and generates the response instruction information. Based on the response instruction information, the target data is retrieved from the business system.

[0054] Once the target data has been retrieved, the instruction execution results and performance metrics are recorded, and an execution log is generated.

[0055] It should be noted that each of the processing modules has an independent instruction control sub-agent. The instruction control sub-agent works in conjunction with the supervisor agent. The supervisor agent generates instructions and assigns tasks, and each instruction control sub-agent receives tasks and executes response instructions, forming a complete and efficient multi-agent collaborative architecture.

[0056] In one specific implementation, the instruction control sub-agent receives and parses the task analysis results to obtain the instruction parameters and execution conditions of the current task. It then performs a rationality verification based on the context and system status, and generates the response instruction information for the digital human to execute. Subsequently, the front-end page receives and parses the response instruction information, verifies the instruction permissions and execution conditions, executes specific interface control operations, updates the interface status in real time, provides visual feedback to the user, records the instruction execution results and performance indicators, and generates an execution log for monitoring and auditing.

[0057] It should be noted that the response instruction information includes the action to be executed, the target object, the parameter configuration, and the expected result, ensuring that the instruction can be executed accurately; the interface control operations include page navigation, element operation, data update, etc., ensuring that user instructions receive accurate and timely responses.

[0058] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0059] Reference Figure 2 This illustration shows an embodiment of an intelligent data retrieval device based on semantic understanding, specifically including the following modules: Specifically, it includes: User instruction preprocessing module 310 is used to generate structured text information based on the current user instruction input by the user in response to the current user instruction. The semantic parsing and task generation module 320 is used to determine semantic information and target resources based on the structured text information, and generate task analysis results and task classification labels based on the semantic information and target resources through a preset supervisory agent. The task analysis results include user intent, instruction entities and contextual relationships. The task allocation module 330 is used to select the corresponding processing module through the supervisory agent based on the task classification label, and transmit the task analysis result to the corresponding processing module. The instruction execution module 340 is used to generate structured response instruction information by controlling the sub-intelligent agent through preset instructions based on the task analysis results, and to call target data based on the response instruction information.

[0060] In one embodiment of the present invention, the user instruction preprocessing module 310 includes: The instruction receiving submodule is used to receive the user instructions issued by the user on the client page; The speech-to-text submodule is used to recognize the user's voice commands and convert them into text command information; The standardization processing submodule is used to perform standardization preprocessing on the current user command according to preset command format rules; The compliance verification submodule is used to perform format verification and security checks on the preprocessed current user command, and generate the corresponding structured text information based on the current user command that conforms to the processing specifications.

[0061] In one embodiment of the present invention, the semantic parsing and task generation module 320 includes: The semantic understanding submodule is used to perform semantic parsing based on the structured text information using the preset large language model, and generate semantic information including the user intent, the instruction entity and the context relationship; The target resource extraction submodule is used to calculate a comprehensive constraint matching score based on the structured text information and a preset target resource library using a multi-constraint weight allocation formula, and to filter and extract the target resources from the target resource library based on the comprehensive constraint matching score. The task generation submodule is used to perform deep analysis through the supervisory agent based on the semantic information and the target resources to generate the task analysis results. The task classification submodule is used to identify task types based on the task analysis results and a predefined task classification system, and generate the task classification labels.

[0062] In one embodiment of the present invention, the task allocation and transmission module 330 includes: The target processing module selection sub-module is used to identify and select the corresponding processing module from the business system based on the task classification label, through the supervisory agent; The routing strategy generation submodule is used to generate a routing strategy based on the task classification tags and the processing module, using a preset intelligent routing model. The task data transmission submodule is used to transmit the task analysis results to the processing module through the supervisory agent according to the routing strategy.

[0063] In one embodiment of the present invention, the instruction execution module 340 includes: The task parsing submodule is used to control the sub-agent to receive and parse the task analysis results through the instruction, and generate the response instruction information. The target data retrieval submodule is used to retrieve target data from the business system based on the response instruction information; The record generation submodule is used to record the instruction execution results and performance indicators, and generate an execution log, after the target data has been called.

[0064] Reference Figure 3This document illustrates a workflow diagram of a semantic understanding-based intelligent data retrieval method according to an embodiment of this application. The workflow steps are as follows: voice input and front-end interaction, voice detection and transcription, semantic granularity and task distribution, instruction recognition and processing, information transmission and communication, and front-end control instruction execution. Through these steps, a fully closed-loop workflow of "instruction input - automatic execution - result presentation" is formed, completely eliminating the need for manual user confirmation. Operational efficiency is significantly improved compared to existing semi-automated solutions. This solves the technical problem that current digital human systems can only achieve simple basic interactions and cannot fully extract, accurately identify, and automatically execute ambiguous instructions and professional terms based on the user's natural language. It establishes an automated mapping relationship between semantic parsing results and business system operations, achieving deep integration of digital human interaction scenarios and business operations. This results in accurate recognition of user instructions, improved user instruction execution response speed, and a user experience that better matches the expectations of human-like interaction. It satisfies the need for efficient interaction between digital humans and users, and constructs an efficient, secure, and convenient intelligent retrieval system for digital human-driven web pages and data charts.

[0065] Reference Figure 4 The illustration shows a computer electronic device for implementing a semantic understanding-based intelligent data retrieval method of the present invention, which may specifically include the following: The aforementioned computer electronic device 1 is manifested in the form of a general-purpose computing device. The components of a computer electronic device 1 may include, but are not limited to: one or more processors or processing units 3, memory 8, and a bus 4 connecting different system components (including memory 8 and processing unit 3).

[0066] Bus 4 represents one or more of several bus architectures, including memory buses or memory controllers, peripheral buses, graphics acceleration ports, processors, or local buses using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Audio / Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0067] A computer electronic device 1 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by a computer electronic device 1, including volatile and non-volatile media, removable and non-removable media.

[0068] Memory 8 may include computer system readable media in the form of volatile memory, such as random access memory 9 and / or cache memory 10. A computer electronic device 1 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 11 may be used to read and write non-removable, non-volatile magnetic media (commonly referred to as a "hard disk drive"). Although Figure 4 As not shown, a disk drive for reading and writing to a removable non-volatile disk (such as a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 4 via one or more data media interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules 13 configured to perform the functions of the embodiments of this application.

[0069] A program / utility 12 having a set (at least one) of program modules 13 may be stored, for example, in memory. Such program modules 13 include—but are not limited to—an operating system, one or more application programs, other program modules 13, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 13 typically perform the functions and / or methods described in the embodiments of this application.

[0070] A computer electronic device 1 can also communicate with one or more external devices 2 (e.g., keyboard, pointing device, display 7, camera, etc.), and with one or more devices that enable an operator to interact with the computer electronic device 1, and / or with any device that enables the computer electronic device 1 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed through I / O interface 6. Furthermore, the computer electronic device 1 can also communicate with one or more networks (e.g., local area network (LAN)), wide area network (WAN), and / or public networks (e.g., the Internet) via network adapter 5. Figure 4 As shown, network adapter 5 communicates with other modules of a computer electronic device 1 via bus 4. It should be understood that, although... Figure 4 As not shown, other hardware and / or software modules may be used in conjunction with a computer electronic device 1, including but not limited to: microcode, device drivers, redundant processing units 3, external disk drive arrays, RAID systems, tape drives, and data backup storage systems 11.

[0071] The processing unit 3 executes various functional applications and data processing by running programs stored in memory 8, such as implementing a method for intelligent data retrieval based on semantic understanding provided in the embodiments of this application.

[0072] That is, when the above-mentioned processing unit 3 executes the above-mentioned program, it realizes: in response to the current user instruction input by the user, it generates structured text information based on the current user instruction; Based on the structured text information, semantic information and target resources are determined, and through a preset supervisory agent, task analysis results and task classification labels are generated based on the semantic information and target resources. The semantic information includes user intent, instruction entities, and contextual relationships. Based on the task classification tags, the supervisory agent selects the corresponding processing module in the business system and transmits the task analysis results to the corresponding processing module. Based on the task analysis results, the sub-agent is controlled by preset instructions to generate structured response instruction information, and the target data is invoked based on the response instruction information.

[0073] In this application embodiment, the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for intelligent data retrieval based on semantic understanding as provided in all embodiments of the present application.

[0074] That is, when the program is executed by the processor, it should implement the following: in response to the current user instruction input by the user, generate structured text information based on the current user instruction; Based on the structured text information, semantic information and target resources are determined, and through a preset supervisory agent, task analysis results and task classification labels are generated based on the semantic information and target resources. The semantic information includes user intent, instruction entities, and contextual relationships. Based on the task classification tags, the supervisory agent selects the corresponding processing module in the business system and transmits the task analysis results to the corresponding processing module. Based on the task analysis results, the sub-agent is controlled by preset instructions to generate structured response instruction information, and the target data is invoked based on the response instruction information.

[0075] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0076] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including—but not limited to—electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0077] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the operator's computer, partially on the operator's computer, as a standalone software package, partially on the operator's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the operator's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider). The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably.

[0078] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0079] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0080] The present application provides a detailed description of a semantic understanding-based intelligent data retrieval method and apparatus. Specific examples have been used to illustrate the principles and implementation methods of the present application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present application. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of the present application. Therefore, the content of this specification should not be construed as a limitation of the present application.

Claims

1. A method for intelligently retrieving data based on semantic understanding, characterized in that, The method is based on digital human technology and a large language model combined with a multi-agent collaborative architecture to perform semantic understanding of user commands. The method includes: In response to the current user instruction input by the user, structured text information is generated based on the current user instruction; Based on the structured text information, semantic information and target resources are determined, and through a preset supervisory agent, task analysis results and task classification tags are generated based on the semantic information and target resources. The semantic information includes user intent, instruction entities, and contextual relationships. Based on the task classification tags, the supervisory agent selects the corresponding processing module in the business system and transmits the task analysis results to the corresponding processing module. Based on the task analysis results, the sub-agent is controlled by preset instructions to generate structured response instruction information, and the target data is invoked based on the response instruction information.

2. The intelligent data retrieval method based on semantic understanding according to claim 1, characterized in that, The step of generating structured text information in response to a current user instruction input by the user includes: The current user instructions are received through the front-end interface, wherein the current user instructions include user voice instructions and text input instructions; The current user command is preprocessed according to preset command format rules; The preprocessed current user instruction is subjected to format verification and security checks, and the corresponding structured text information is generated based on the current user instruction that conforms to the processing specifications.

3. The intelligent data retrieval method based on semantic understanding according to claim 1, characterized in that, The step of determining semantic information and target resources based on the structured text information, and generating task analysis results and task classification labels based on the semantic information and target resources through a preset supervisory agent, wherein the semantic information includes user intent, instruction entities, and contextual relationships, includes: Based on the structured text information, semantic parsing is performed using the preset large language model to generate semantic information including the user intent, the instruction entity, and the context relationship; Based on the structured text information and the preset target resource library, a comprehensive constraint matching score is calculated using a multi-constraint weight allocation formula. The target resources are then selected and extracted from the target resource library based on the comprehensive constraint matching score. Based on the semantic information and the target resources, the supervisory agent generates the task analysis results and the task classification labels.

4. The intelligent data retrieval method based on semantic understanding according to claim 3, characterized in that, The step of calculating a comprehensive constraint matching score based on the structured text information and a preset target resource library using a multi-constraint weight allocation formula, and then filtering and extracting the target resources from the target resource library based on the comprehensive constraint matching score, includes: Based on the structured text information and the preset target resource library, the time constraint matching degree, location constraint matching degree and data type constraint matching degree are obtained respectively; Based on the time constraint matching degree, the location constraint matching degree, and the data type constraint matching degree, the comprehensive constraint matching score is calculated using the multi-constraint weight allocation formula; Based on the comprehensive constraint matching score, the target resources are selected and extracted from the target resource library.

5. The intelligent data retrieval method based on semantic understanding according to claim 3, characterized in that, The step of generating the task analysis result and the task classification label by the supervisory agent based on the semantic information and the target resource includes: Based on the semantic information and the target resources, the supervisory agent performs deep analysis to generate the task analysis results; Based on the task analysis results and the predefined task classification system, task types are identified, and task classification labels are generated.

6. The intelligent data retrieval method based on semantic understanding according to claim 1, characterized in that, The step of selecting the corresponding processing module in the business system based on the task classification label by the supervisory agent, and transmitting the task analysis results to the corresponding processing module, includes: Based on the task classification tags, the supervisory agent identifies and selects the corresponding processing module from the business system. Based on the task classification tags and the processing module, a routing strategy is generated through a preset intelligent routing model; According to the routing strategy, the task analysis results are transmitted to the processing module through the supervisory agent.

7. The intelligent data retrieval method based on semantic understanding according to claim 1, characterized in that, The step of generating structured response instruction information by controlling a sub-agent through preset instructions based on the task analysis results, and calling target data based on the response instruction information, includes: The sub-agent receives and parses the task analysis results through the instruction control, and generates the response instruction information. Based on the response instruction information, retrieve the target data from the business system; Once the target data has been retrieved, the instruction execution results and performance metrics are recorded, and an execution log is generated.

8. A semantic understanding-based intelligent data retrieval device, characterized in that, include: The user instruction preprocessing module is used to generate structured text information based on the current user instruction input by the user in response to the current user instruction. The semantic parsing and task generation module is used to determine semantic information and target resources based on the structured text information, and generate task analysis results and task classification tags based on the semantic information and target resources through a preset supervisory agent. The task analysis results include user intent, instruction entities and contextual relationships. The task allocation and transmission module is used to select the corresponding processing module through the supervisory agent based on the task classification label, and transmit the task analysis results to the corresponding processing module. The instruction execution module is used to generate structured response instruction information for the sub-intelligent agent through preset instructions based on the task analysis results, and to call target data based on the response instruction information.

9. A computer electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein when executed by the processor, the computer program implements the steps of the semantic understanding-based intelligent data retrieval method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the steps of the semantic understanding-based intelligent data retrieval method as described in any one of claims 1 to 7.

Images