CN-122021655-A - Big data direct-driven application generation and adaptation method of semantic reasoning engine

CN122021655ACN 122021655 ACN122021655 ACN 122021655ACN-122021655-A

Abstract

The invention provides a big data direct-driven application generation and adaptation method of a semantic reasoning engine, which relates to the technical field of artificial intelligence and comprises the steps of analyzing a big data source and constructing a data semantic graph through an AI semantic reasoning engine; and selecting and instantiating a component template according to the semantic attribute of the binding node to generate the executable application. During application running, the change of a data source is monitored, the map is updated in an increment mode, and affected component examples are positioned and partially reconstructed through difference vectors, so that automatic synchronization and adaptation of application and data change are achieved.

Inventors

SUN SHUMENG
Jia Songrui
YAN MIN
ZHANG RUIDONG
ZHANG HAORAN

Assignees

北京亦庄智能城市研究院集团有限公司

Dates

Publication Date: 20260512
Application Date: 20260416

Claims (9)

1. The big data direct-driven application generation and adaptation method of the semantic reasoning engine is characterized by comprising the following steps: Carrying out deep semantic analysis on a large data source through an AI semantic reasoning engine, extracting semantic elements and constructing a data semantic graph, wherein the data semantic graph takes data entities as nodes and the association between the entities as edges, and the nodes and the edges carry semantic attributes; Deducing an application component structure based on the semantic features of the data semantic graph, and establishing a direct binding relation between each application component in the application component structure and a corresponding node or edge in the data semantic graph; Selecting a component template from a parameterized component template library according to semantic attributes of nodes or edges directly bound with each application component in the application component structure, and directly instantiating variable parameters of the component template by utilizing the semantic attributes to generate executable applications; During the running of the executable application, monitoring the change of the big data source, performing incremental update on the data semantic graphs through an AI semantic reasoning engine, and calculating a difference vector; and positioning the affected component instance in the executable application according to the difference vector and the direct binding relation, and executing local reconstruction on the component instance based on the updated data semantic graph.
2. The method of claim 1, wherein the step of performing deep semantic parsing on the large data source by the AI semantic reasoning engine, extracting semantic elements and constructing a data semantic graph, the data semantic graph having data entities as nodes and inter-entity associations as edges, the nodes and edges carrying semantic attributes comprises: the AI semantic reasoning engine performs unified analysis on structured data and unstructured data in a big data source to identify a data entity and attribute information thereof; analyzing reference dependence, including subordinate and interactive association among different data entities in the big data source, and establishing association semantics among the entities; Mapping the data entities into nodes of data semantic graphs, mapping the associated semantics among the entities into edges of the data semantic graphs, and respectively marking the entity type identifiers, the entity feature descriptions, the entity constraint conditions and the associated semantics as semantic attributes on the corresponding nodes and edges; and executing semantic reasoning based on rule reasoning and mode reasoning on the data semantic graph, deducing an implicit entity association and attribute transfer relation according to the annotated semantic attribute, and supplementing the implicit entity association and attribute transfer relation to the data semantic graph to form a complete data semantic graph.
3. The method of claim 1, wherein deriving an application component structure based on semantic features of the data semantic graph, the step of establishing a direct binding relationship between each application component in the application component structure and a corresponding node or edge in the data semantic graph comprises: analyzing topological structure features and semantic attribute distribution features of the data semantic graphs, determining the hierarchy level of the corresponding application component according to the connectivity of the nodes and the complexity of the semantic attributes, determining the interaction logic of the corresponding application component according to the dependence of the association direction of the edges and the semantic attributes, and generating an application component structure with a hierarchy structure and interaction relation; and establishing a direct binding relation comprising a semantic attribute access path between the component identifier of the application component and the map identifier of the corresponding node or side in the data semantic map for each application component in the application component structure.
4. The method of claim 1, wherein for each application component in the application component structure, the step of selecting a component template from a library of parameterized component templates based on semantic attributes of its directly bound nodes or edges, and directly instantiating variable parameters of the component template with the semantic attributes, the step of generating an executable application comprises: For each application component in the application component structure, carrying out semantic reasoning on semantic attributes of nodes or edges directly bound with the application component, deducing technical specification requirements of the application component according to semantic constraint relations in the semantic attributes, calculating semantic adaptation degree of each candidate component template in a parameterized component template library according to performance indexes and interaction characteristics in the technical specification requirements, and selecting a component template with highest semantic adaptation degree; extracting semantic definitions of variable parameters in the selected component templates, converting performance indexes and interaction characteristics in the technical specification requirements into parameter value spaces corresponding to the semantic definitions of the variable parameters, determining parameter values meeting the semantic attribute constraint from the parameter value spaces, assigning the determined parameter values to the variable parameters to complete instantiation of the component templates to generate component instances, and assembling all the instantiated component instances according to the application component structure to generate executable applications.
5. The method of claim 4, wherein deriving a technical specification requirement for the application component according to the semantic constraint relationship in the semantic attribute, and calculating the semantic fitness of each candidate component template in the parameterized component template library according to the performance index and the interaction characteristic in the technical specification requirement, and selecting the component template with the highest semantic fitness comprises: extracting semantic attribute history change records of nodes or edges corresponding to the technical specification requirements, and analyzing change trends and change periods of the semantic attribute history change records to establish a semantic evolution rule; Deducing potential semantic expansion requirements of the node or the edge in a preset evolution time window through the semantic evolution rule, and generating an expansion technical specification requirement comprising the current technical specification requirement and the potential semantic expansion requirement; The method comprises the steps of calculating the instant adaptation degree of the current technical specification requirement and the coverage rate of a variable parameter configuration space on the potential semantic expansion requirement as expansion adaptation degree according to each candidate component template in a parameterized component template library, determining the comprehensive semantic adaptation degree according to the weighted combination of the instant adaptation degree and the expansion adaptation degree, selecting a component template with the highest comprehensive semantic adaptation degree, reserving a parameter expansion interval for variable parameters corresponding to the potential semantic expansion requirement during instantiation, and completing the adaptation by component instances through parameter value adjustment in the parameter expansion interval when the potential semantic expansion requirement is actually triggered during operation.
6. The method of claim 1, wherein the step of computing a disparity vector by incrementally updating the data semantic graph by an AI semantic reasoning engine comprises: Analyzing the influence degree of data change on the semantic attributes of nodes and edges in the data semantic graphs by the AI semantic reasoning engine, and identifying the nodes and edges with the influence degree exceeding a preset influence threshold as direct semantic change elements; According to semantic dependency relationships in the data semantic graphs, propagating semantic changes of the direct semantic change elements, dynamically judging whether the semantic change degree of nodes and edges affected by propagation exceeds a continuous propagation threshold in the propagation process, marking the nodes and edges exceeding the continuous propagation threshold as indirect semantic change elements, and performing incremental update on the semantic attributes of the direct semantic change elements and the indirect semantic change elements; and calculating a change vector of semantic attribute values before and after updating in a semantic space aiming at each node and edge which completes incremental updating, and marking a change confidence coefficient and an influence weight for the change vector by combining constraint rules of the semantic attributes and topological positions of the nodes or edges in the data semantic graph to generate a difference vector comprising a semantic change direction, the change confidence coefficient and the influence weight.
7. The method of claim 1, wherein locating the affected component instance in the executable application according to the disparity vector and the direct binding relationship, the step of performing local reconstruction on the component instance based on the updated data semantic graph comprises: Extracting a semantic dimension identifier which changes from the difference vector, acquiring a component instance which is bound with a graph node corresponding to the semantic dimension which changes according to the direct binding relation, analyzing a code structure and data acquisition logic of the component instance, and identifying the semantic dimension which is actually used in the component instance; Matching the changed semantic dimension identification with the actually used semantic dimension, extracting influence weight in the difference vector from the semantic dimension which is successfully matched, and screening component instances with the influence weight exceeding a reconstruction threshold as affected component instances; For each affected component instance, locating a code segment using the semantic dimension of successful matching based on the code structure and data acquisition logic, dividing the component instance into a to-be-reconstructed part containing the code segment and a reserved part not containing the code segment, regenerating codes for the to-be-reconstructed part based on the updated data semantic graph, keeping the codes of the reserved part unchanged, and fusing the regenerated codes with the codes of the reserved part to complete local reconstruction.
8. An electronic device, comprising: A processor; A memory for storing processor-executable instructions; Wherein the processor is configured to invoke the instructions stored in the memory to perform the method of any of claims 1 to 7.
9. A computer readable storage medium having stored thereon computer program instructions, which when executed by a processor, implement the method of any of claims 1 to 7.

Description

Big data direct-driven application generation and adaptation method of semantic reasoning engine Technical Field The invention relates to an artificial intelligence technology, in particular to a big data direct-driven application generation and adaptation method of a semantic reasoning engine. Background In the prior art, big data driven application development generally follows a traditional flow. The development team needs to analyze the business requirements first, then data is extracted, cleaned and converted from the big data source by the data engineer to form a data model or data interface that can be used for application. Based on these data models, application developers manually design application architecture, write business logic code, and build user interfaces. After the application deployment is online, if the structure or content of the underlying data source is changed, for example, the fields of the data table are added and deleted, the business rule is adjusted or the data relationship is changed, a developer is often required to manually analyze the change influence range, manually modify the corresponding data interface, business logic and even the front-end display code, and then test and deploy again. This conventional approach has significant drawbacks. The whole application development process is highly dependent on manpower, the links from data understanding to code realization are numerous, the development period is long and the cost is high. A more prominent problem is the contradiction between the static nature of the application and the data dynamics. Data sources in large data environments often evolve and change, while traditional applications are loosely associated with data sources indirectly through hard-coding or configuration files. When the data changes, an effective automatic mechanism is lacking to sense the concrete semantic meaning of the change, and the part to be adjusted in the application is accurately positioned, so that the application is difficult to maintain and slow in response, the real-time synchronization and adaptation with the dynamic data environment are difficult to realize, and the vitality and flexibility of the application are severely restricted. Disclosure of Invention The embodiment of the invention provides a big data direct-driven application generation and adaptation method of a semantic reasoning engine, which can solve the problems in the prior art. In a first aspect of the embodiment of the present invention, a big data direct-driven application generating and adapting method of a semantic reasoning engine is provided, including: Carrying out deep semantic analysis on a large data source through an AI semantic reasoning engine, extracting semantic elements and constructing a data semantic graph, wherein the data semantic graph takes data entities as nodes and the association between the entities as edges, and the nodes and the edges carry semantic attributes; Deducing an application component structure based on the semantic features of the data semantic graph, and establishing a direct binding relation between each application component in the application component structure and a corresponding node or edge in the data semantic graph; Selecting a component template from a parameterized component template library according to semantic attributes of nodes or edges directly bound with each application component in the application component structure, and directly instantiating variable parameters of the component template by utilizing the semantic attributes to generate executable applications; During the running of the executable application, monitoring the change of the big data source, performing incremental update on the data semantic graphs through an AI semantic reasoning engine, and calculating a difference vector; and positioning the affected component instance in the executable application according to the difference vector and the direct binding relation, and executing local reconstruction on the component instance based on the updated data semantic graph. Carrying out deep semantic analysis on a large data source through an AI semantic reasoning engine, extracting semantic elements and constructing a data semantic graph, wherein the data semantic graph takes data entities as nodes and the association between the entities as edges, and the steps of carrying semantic attributes by the nodes and the edges comprise: the AI semantic reasoning engine performs unified analysis on structured data and unstructured data in a big data source to identify a data entity and attribute information thereof; analyzing reference dependence, including subordinate and interactive association among different data entities in the big data source, and establishing association semantics among the entities; Mapping the data entities into nodes of data semantic graphs, mapping the associated semantics among the entities into edges of the data semantic graph