CN-122019792-A - Knowledge graph guided intelligent agent data full life cycle management method and device

CN122019792ACN 122019792 ACN122019792 ACN 122019792ACN-122019792-A

Abstract

The application discloses a knowledge graph guided intelligent agent data full life cycle management method and device, and relates to the technical field of intelligent agents. The method comprises the steps of connecting a data management platform to an intelligent agent of a client and constructing a core knowledge graph, conducting intelligent guidance by using the core knowledge graph according to task requests in a data acquisition stage, conducting data acquisition, conducting intelligent treatment on a plurality of original data by using the core knowledge graph in a data preprocessing stage, conducting intelligent organization on a plurality of preprocessed data by using the core knowledge graph in a data storage stage, conducting intelligent service by using an updated core knowledge graph according to task information in a data application stage, and conducting intelligent treatment on a secondary updated core knowledge graph in a data archiving stage. The problems of high data acquisition blindness, high data management cost, data organization confusion, data application and data management disjoint and data archiving strategy rigidification in the prior art are solved.

Inventors

LI YINGBIN
ZHANG BING
WANG WENDI
SHI CHUNHUA

Assignees

海天地数码科技（北京）有限公司

Dates

Publication Date: 20260512
Application Date: 20260407

Claims (10)

1. A knowledge-graph-guided agent data full life cycle management method, characterized in that the method comprises: Connecting a data management platform to an intelligent agent of a client, and constructing a corresponding core knowledge graph in a cloud server according to metadata of the intelligent agent; In the data acquisition stage, according to the task request of an intelligent agent, intelligent guidance is carried out by using a core knowledge graph, and data acquisition is carried out in a data source to obtain a plurality of original data; in the data preprocessing stage, a core knowledge graph is used for intelligently managing a plurality of original data to obtain data after intervention processing; In the data storage stage, using a core knowledge graph to intelligently organize a plurality of preprocessed data to obtain an updated core knowledge graph; in the data application stage, according to the task information of the intelligent agent, using the updated core knowledge graph to carry out intelligent service, providing task support for the intelligent agent, and recording the data use behavior of the intelligent agent to obtain a secondary updated core knowledge graph; And in the data archiving stage, performing intelligent treatment on the secondary updated core knowledge graph, and eliminating low-value data entities to obtain the tertiary updated core knowledge graph.
2. The knowledge-graph-guided agent data full life cycle management method of claim 1, wherein connecting a data management platform to an agent at a client, and constructing a corresponding core knowledge graph in a cloud server according to application scenario information of the agent, comprises: logging in a data management platform at a client, and connecting the data management platform to an agent locally stored at the client; Collecting metadata of the intelligent agent by using a data management platform, and uploading the metadata to a cloud server through an encryption security channel; And constructing a corresponding core knowledge graph in the cloud server according to the metadata of the intelligent agent.
3. The knowledge-graph-guided whole life cycle management method of intelligent agent data according to claim 2, wherein constructing a corresponding core knowledge graph in a cloud server according to metadata of intelligent agent comprises: in a cloud server, defining a multi-dimensional ontology layer of a core knowledge graph; extracting structured data, semi-structured data and unstructured data in metadata of an intelligent agent, and filling the structured data, the semi-structured data and unstructured data into a multi-dimensional body layer to obtain an initial core knowledge graph; and constructing a dynamic evolution mechanism for the initial core knowledge graph to obtain a final core knowledge graph.
4. The knowledge-graph-guided agent data full life cycle management method of claim 3, wherein in the data collection stage, according to the task request of the agent, the intelligent guidance is performed by using the core knowledge graph, the data collection is performed in the data source, and a plurality of original data are obtained, including: In the data acquisition stage, a data management platform is used for acquiring a task request of an intelligent agent, and the task request is uploaded to a cloud server through an encryption security channel; In a cloud server, analyzing a task request by using a pre-trained entity identification model to obtain a request entity, and inquiring a data entity and an intelligent entity associated with the request entity in a core knowledge graph to obtain a data requirement; inquiring a data source network related to the data requirement in a core knowledge graph, and searching an optimal data acquisition path combination in the data source network by using an improved optimizing algorithm; and according to the optimal data acquisition path combination, a cloud server is used for carrying out data acquisition in different data sources, so as to obtain a plurality of original data.
5. The knowledge-based intelligent agent data full life cycle management method of claim 4, wherein querying a data source network associated with the data requirement in a core knowledge-based graph and using an improved optimization algorithm, searching for an optimal data acquisition path combination in the data source network, comprising: Querying a data source network associated with the data requirement in a core knowledge graph, wherein the data source network comprises a plurality of data sources corresponding to a plurality of data items to be acquired in the data requirement; The data acquisition path combination of a plurality of data sources is encoded into individual vectors of ISBO algorithm, and ISBO population parameters and maximum iteration times are set; Initializing by using a Tent chaotic mapping sequence according to ISBO population parameters to obtain an initial ISBO population, wherein each ISBO individual in the ISBO population corresponds to an alternative data acquisition path combination; acquiring an fitness value of each initial ISBO individual by using a fitness function, and taking the initial ISBO individual with the optimal fitness value as an optimal solution; Introducing a Levy flight strategy and a convergence factor, and constructing a pavilion for searching or developing a decorative pavilion for the initial ISBO population with searching probability to obtain an updated first ISBO population; randomly selecting a plurality of first ISBO individuals from the updated first ISBO population according to the theft probability to simulate the theft behavior to obtain a plurality of updated second ISBO individuals; introducing a dynamic reverse mechanism, and randomly selecting a plurality of first ISBO individuals from the updated first ISBO population to conduct direction according to reverse probability to obtain a plurality of updated third ISBO individuals; Acquiring fitness values of each updated first, second and third ISBO individuals by using a fitness function, and updating the updated ISBO individuals with optimal fitness values into optimal solutions; when the iteration number reaches the maximum iteration number or the fitness value of the optimal solution meets the requirement, stopping iteration update of ISBO population, and outputting the optimal solution of the current iteration; and decoding individual vectors of ISBO individuals corresponding to the optimal solution to obtain an optimal data acquisition path combination in the data source network.
6. The knowledge-graph-guided agent data full life cycle management method of claim 5, wherein in the data preprocessing stage, the core knowledge graph is used to intelligently manage a plurality of original data to obtain the data after the intervention processing, including: In the data preprocessing stage, entity extraction is carried out on each original data by using an entity identification model to obtain a plurality of material entities, and the material entities are associated with the corresponding data entities in the core knowledge graph to obtain a plurality of associated data entities; according to a data element model and a cleaning rule defined in the core knowledge graph, cleaning, de-duplicating and formatting a plurality of original data provided with the associated data entity to obtain a plurality of standard data provided with the associated data entity; And according to the domain knowledge defined in the core knowledge graph, carrying out semantic annotation and entity linking on all related data entities of a plurality of standard data to obtain a plurality of preprocessed data provided with annotation information.
7. The knowledge-graph-guided agent data full life cycle management method of claim 6, wherein in the data storage stage, the core knowledge graph is used to intelligently organize the plurality of preprocessed data to obtain the updated core knowledge graph, comprising: In the data storage stage, the associated data entity and the labeling information of each preprocessed data are used as new data entities, and updated to a core knowledge graph; According to the graph structure of the core knowledge graph, establishing semantic association between the preprocessed data, and vectorizing each preprocessed data to obtain a semantic index; and adding the semanteme index to a new data entity corresponding to the core knowledge graph to obtain an updated core knowledge graph.
8. The knowledge-graph-guided whole life cycle management method of intelligent agent data according to claim 7, wherein in the data application stage, according to the task information of the intelligent agent, using the updated core knowledge graph to perform intelligent service, providing task support for the intelligent agent, and recording the data usage behavior of the intelligent agent to obtain a secondary updated core knowledge graph, comprising: In the data application stage, a data management platform is used for collecting task information of an intelligent agent, and the task information is uploaded to a cloud server through an encryption security channel; In a cloud server, performing entity extraction on task information by using an entity identification model to obtain a plurality of task entities, and inquiring a data entity and a knowledge path associated with the task entities in a core knowledge graph to obtain a knowledge subgraph; according to the knowledge subgraph and the historical feedback information, performing decision generation by using a pre-constructed intelligent service decision model to obtain an intelligent service decision, and sending the knowledge subgraph and the intelligent service decision to an agent at a client through a data management platform to provide task support for the agent; recording the data use behavior of the intelligent agent on the knowledge subgraph on the basis of intelligent service decision by using a data management platform, and uploading the data use behavior to a cloud server through an encryption security channel; And in the cloud server, taking the data use behavior as a new entity relation of all data entities in the knowledge subgraph, and adding the new entity relation to the corresponding data entities in the core knowledge graph to obtain a secondary updated core knowledge graph.
9. The knowledge-graph-guided agent data full life cycle management method of claim 8, wherein in a data archiving stage, performing intelligent treatment on the secondary updated core knowledge graph, eliminating low-value data entities, and obtaining a tertiary updated core knowledge graph, comprising: in the data archiving stage, collecting the value evaluation basis of all data entities in the secondarily updated core knowledge graph; According to the value evaluation basis, performing value prediction by using a pre-constructed value prediction model to obtain the future value of each data entity; And according to a preset archiving strategy and compliance requirements, archiving and destroying the data entities with future values lower than the value threshold, and updating the states of the corresponding data entities in the secondarily updated core knowledge graph to obtain the thirdly updated core knowledge graph.
10. A knowledge-graph-guided agent data full life cycle management apparatus for implementing the agent data full life cycle management method according to any one of claims 1 to 9, the apparatus comprising: The knowledge graph construction unit is used for connecting the data management platform to the intelligent agent of the client and constructing a corresponding core knowledge graph in the cloud server according to the metadata of the intelligent agent; The data acquisition unit is used for performing intelligent guidance by using a core knowledge graph according to the task request of the intelligent agent in the data acquisition stage, and performing data acquisition in a data source to obtain a plurality of original data; The data preprocessing unit is used for intelligently managing a plurality of original data by using a core knowledge graph in a data preprocessing stage to obtain data after intervention processing; the data storage unit is used for intelligently organizing the plurality of preprocessed data by using the core knowledge graph in the data storage stage to obtain an updated core knowledge graph; The data application unit is used for performing intelligent service by using the updated core knowledge graph according to the task information of the intelligent agent in the data application stage, providing task support for the intelligent agent, and recording the data use behavior of the intelligent agent to obtain a secondary updated core knowledge graph; And the data archiving unit is used for intelligently processing the secondary updated core knowledge graph in the data archiving stage, eliminating the low-value data entity and obtaining the tertiary updated core knowledge graph.

Description

Knowledge graph guided intelligent agent data full life cycle management method and device Technical Field The invention relates to the technical field of intelligent agents, in particular to a knowledge graph guided intelligent agent data full life cycle management method and device. Background With the rapid development of artificial intelligence technology, various intelligent agents (such as intelligent customer service, automatic driving system, personalized recommendation engine, industrial robot, etc.) have been widely used in various aspects of social production and life. The performance and intelligence level of these agents is highly dependent on high quality, large scale, diverse data support. However, the data generated by the intelligent agent in the operation process has the characteristics of wide sources, complex structure, dynamic evolution, uneven value density and the like, and brings great challenges to data management. Traditional data management methods typically employ a "chimney" or islanding management mode, optimized for a particular stage of the data lifecycle (e.g., acquisition, storage, analysis), lacking in global and coherency. This results in a series of problems: 1) The blindness of data acquisition is high, and the dynamic requirements of the intelligent agent are difficult to accurately match; 2) The data management cost is high, a large number of manual rules and scripts are relied on, the efficiency is low, and errors are easy to occur; 3) The data organization is chaotic, the data lacks effective semantic association, and cross-source and cross-domain knowledge discovery and multiplexing are difficult to realize; 4) The data application is disjointed with the data management, and the intelligent agent data cannot be dynamically evaluated and optimized according to the actual use behaviors of the intelligent agent; 5) The data archiving strategy is stiff, low-value data cannot be effectively identified and handled, and storage resource waste and data asset bloat are caused. Therefore, how to provide a scheme capable of realizing full life cycle intelligence, automation and semantic management of data, and to penetrate and guide the full life cycle management flow of intelligent agent data from collection, management, storage and application to archiving, so as to improve the utilization efficiency and value of intelligent agent data, and support continuous learning and evolution of intelligent agent, is a technical problem to be solved currently. Disclosure of Invention The invention provides a knowledge graph guided intelligent agent data full life cycle management method and device, which solve the problems of high data acquisition blindness, high data management cost, confusion of data organization, disconnection of data application and data management and rigidification of a data archiving strategy in the prior art. In a first aspect, an embodiment of the present invention provides a knowledge-graph-guided agent data full life cycle management method, where the method includes: Connecting a data management platform to an intelligent agent of a client, and constructing a corresponding core knowledge graph in a cloud server according to metadata of the intelligent agent; In the data acquisition stage, according to the task request of an intelligent agent, intelligent guidance is carried out by using a core knowledge graph, and data acquisition is carried out in a data source to obtain a plurality of original data; in the data preprocessing stage, a core knowledge graph is used for intelligently managing a plurality of original data to obtain data after intervention processing; In the data storage stage, using a core knowledge graph to intelligently organize a plurality of preprocessed data to obtain an updated core knowledge graph; in the data application stage, according to the task information of the intelligent agent, using the updated core knowledge graph to carry out intelligent service, providing task support for the intelligent agent, and recording the data use behavior of the intelligent agent to obtain a secondary updated core knowledge graph; And in the data archiving stage, performing intelligent treatment on the secondary updated core knowledge graph, and eliminating low-value data entities to obtain the tertiary updated core knowledge graph. The technical scheme provided by the embodiment of the application at least has the following beneficial effects: The intelligent management system takes a knowledge graph as a core engine for the first time, and forms a closed loop management system of data driving, knowledge guiding and self-optimization throughout the whole life cycle of data acquisition, management, storage and application and archiving, breaks the phase barriers of traditional intelligent agent data management, utilizes the knowledge graph to understand task demands in the data acquisition phase and combines an improved optimizing algorithm to pla