Chart-interactive knowledge graph query system and method

Abstract

The application relates to a knowledge graph query system and method based on a graph database and a document type database and based on graph linkage, and belongs to the technical field of knowledge graph retrieval. The system comprises an interactive multidimensional visualization display module of a domain knowledge model and an interactive knowledge reasoning and exploration module. The interactive multidimensional visualization display module is used for querying the structural relationship between entities, comprises a force-directed layout view, a hierarchical layout view and a matrix layout view, and further comprises a GIS geographic view, a hierarchical progressive view, a graph structure view, a table structure view, a microscale view and a macroscale view. The interactive knowledge reasoning and exploration module is used for accurately positioning data information, and the data query mode provided by the interactive knowledge reasoning and exploration module comprises global filtering, local filtering and ontology navigation. The system and method provided by the application can better display and query massive data when facing a knowledge graph with massive data, so that the required information of a user can be efficiently and accurately found.

Description

Technical Field

[0001] This invention belongs to the field of knowledge graph retrieval technology, specifically relating to a knowledge graph query system and method based on graph database and document database with graph linkage. Background Technology

[0002] With the development of knowledge graphs, their data scale is growing exponentially. Tencent's TopBase already has over 100 million entities and 2.2 billion triples; Google's knowledge graph has reached 1 billion entities, and Baidu's has reached 5 billion entities. Traditional relational databases experience a sharp decline in subgraph retrieval performance as data size increases, while graph databases perform poorly in attribute retrieval. Faced with gigabytes or even petabytes of data, knowledge graphs place higher demands on data retrieval; traditional query methods relying solely on graph or relational databases are no longer sufficient to meet these needs. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a knowledge graph query system and method based on graph databases and document databases that link graphs and documents, enabling the knowledge graph to retrieve the required data with high performance and intuitiveness even when faced with data at the GB or even PB level.

[0004] To achieve the above objectives, the present invention adopts the following technical solution:

[0005] A knowledge graph query system based on graph databases and document databases with interconnected graphs and documents includes an interactive multidimensional visualization module for domain knowledge models and an interactive knowledge reasoning and exploration module for domain knowledge models.

[0006] The interactive multidimensional visualization module of the domain knowledge model is used to query the structural relationships between entities and provides multiple ways to view the domain knowledge model, including force-guided layout view, hierarchical layout view and matrix layout view;

[0007] The interactive multidimensional visualization module of the domain knowledge model also provides multiple perspectives for viewing the domain knowledge model, including GIS geographic view, hierarchical progressive view, graph structure view, table structure view, micro-scale view and macro-scale view;

[0008] The interactive knowledge reasoning and exploration module of the domain knowledge model is used to accurately locate data information, and the data exploration methods provided include global filtering, local filtering and ontology navigation.

[0009] Furthermore, in the knowledge graph query system based on graph databases and document databases, as described above, the force-guided layout view is used to display ordinary physical structural relationships, the hierarchical layout view is used to display clear hierarchical relationships between entities, and the matrix layout view is used to display relationships between large-scale entities.

[0010] Furthermore, in the knowledge graph query system based on graph databases and document databases, as described above, the GIS geographic view is used to display entities with latitude and longitude data. The GIS geographic view includes a tree view to display the conceptual model, attribute model, label model, and relation model of the domain knowledge model.

[0011] Furthermore, in the knowledge graph query system based on graph databases and document databases, as described above, the hierarchical progressive view is oriented towards business logic and is used to trace the origins of entity relationships.

[0012] Furthermore, in the knowledge graph query system based on graph databases and document databases, as described above, the graph structure view is used to display the relationships between entities; the table structure view is used to display detailed attribute information of entities and relationships.

[0013] Furthermore, in the knowledge graph query system based on graph databases and document databases, as described above, the micro-scale view is used by users to view all nodes; the macro-scale view is used by users to view some nodes in clusters, and the clustering methods include clustering of nodes of the same type, clustering of related nodes, and clustering of multiple selected nodes.

[0014] Furthermore, in the knowledge graph query system based on graph databases and document databases, as described above, the global filtering method is used to retrieve globally matching data. It performs precise conditional retrieval of the entire knowledge graph based on the ontology model, specifically including the following steps:

[0015] S11, Global Selection Concept;

[0016] S12. Select the relationships related to the concept in the knowledge graph ontology model;

[0017] S13. Set the filtering conditions for the attributes of the concept and the relationship;

[0018] S14. Filter the concepts and relationships in the document database to select entities and edges that meet the filtering conditions;

[0019] S15. In the graph database, select triples that meet the requirements by type and graph structure, and combine them with the entities and edges that meet the requirements queried in step S14 to select data that meets the requirements.

[0020] Furthermore, in the knowledge graph query system based on graph databases and document databases, as described above, the local filtering method uses entities as the starting point to perform precise conditional indexing on data associated with those entities, specifically including the following steps:

[0021] S21. Select an entity in the workspace;

[0022] S22. Select local filtering, and then select the relationship related to the concept to which the entity belongs in the knowledge graph ontology model;

[0023] S23. Set the filtering conditions for the attributes of the concept and the relationship;

[0024] S24. Filter the concepts and relationships in the document database to select entities and edges that meet the filtering conditions;

[0025] S25. In the graph database, select triples that meet the requirements by type and graph structure, and combine them with the entities and edges that meet the requirements queried in step S24 to select data that meets the requirements.

[0026] Furthermore, in the knowledge graph query system based on graph databases and document databases, as described above, the ontology navigation method performs coarse-grained data retrieval and filtering starting from entity nodes, specifically including the following steps:

[0027] S31. Select entity;

[0028] S32. Select the relationships to be expanded from the relationships related to the entity.

[0029] The knowledge graph query system and method based on graph database and document database linkage provided by this invention has the following significant technical effects:

[0030] The system and method described in this invention, by employing interactive multidimensional visualization technology and interactive knowledge reasoning and exploration technology of domain knowledge models, can better display and query massive amounts of data when faced with knowledge graphs containing massive amounts of data. This enables users to efficiently and accurately find the information they need, clarify data relationships, assist users in decision-making, reduce the time users spend querying and processing data, and maximize the usability of massive data knowledge graphs. Attached Figure Description

[0031] Figure 1 It is the force guide view in the interactive multidimensional visualization module of the domain knowledge model described in this invention;

[0032] Figure 2It is the hierarchical view in the interactive multidimensional visualization module of the domain knowledge model described in this invention;

[0033] Figure 3 It is the matrix view in the interactive multidimensional visualization module of the domain knowledge model described in this invention;

[0034] Figure 4 It is the GIS view in the interactive multidimensional visualization module of the domain knowledge model described in this invention;

[0035] Figure 5 It is a hierarchical view in the interactive multidimensional visualization module of the domain knowledge model described in this invention;

[0036] Figure 6 It is the graph structure view in the interactive multidimensional visualization module of the domain knowledge model described in this invention;

[0037] Figure 7 It is a table structure view in the interactive multidimensional visualization module of the domain knowledge model described in this invention;

[0038] Figure 8 This is a microscale view in the interactive multidimensional visualization module of the domain knowledge model described in this invention;

[0039] Figure 9 It is the macroscopic view in the interactive multidimensional visualization module of the domain knowledge model described in this invention;

[0040] Figure 10 It is the global filter view in the interactive knowledge reasoning and exploration module of the domain knowledge model described in this invention;

[0041] Figure 11 This is a local filtered view in the interactive knowledge reasoning and exploration module of the domain knowledge model described in this invention;

[0042] Figure 12 It is the ontology navigation view in the interactive knowledge reasoning and exploration module of the domain knowledge model described in this invention;

[0043] Figure 13 This is a flowchart of the global filtering method in the interactive knowledge reasoning and exploration module of the domain knowledge model described in this invention;

[0044] Figure 14 This is a flowchart of the local filtering method in the interactive knowledge reasoning and exploration module of the domain knowledge model described in this invention;

[0045] Figure 15 The flowchart of the ontology navigation method in the interactive knowledge reasoning and exploration module of the domain knowledge model described in this invention. Detailed Implementation

[0046] The query system and method for knowledge graph graph linkage based on graph database and document database provided by this invention can be used for large-scale and efficient data query in distributed knowledge graph systems. The following describes the implementation of this invention in further detail with application examples of various query methods and in conjunction with the accompanying drawings.

[0047] The system provided by this invention includes an interactive multidimensional visualization module for a domain knowledge model, and an interactive knowledge reasoning and exploration module for a domain knowledge model. The structure and function of these two modules are described in detail below.

[0048] 1. Interactive multi-dimensional visualization module for domain knowledge models

[0049] To meet diverse visualization needs, the domain knowledge model offers multi-dimensional visualization methods. This module provides a relationship view, primarily used to query structural relationships between entities. The relationship view includes various layout options such as force-guided layout, hierarchical layout, and matrix layout to display different structural relationships.

[0050] See Figure 1 As shown, the force-guided layout demonstrates the physical layout method. Users can adjust the physical layout distance by adjusting the physical parameters. It is mainly used to demonstrate ordinary structural relationships.

[0051] See Figure 2 As shown, the hierarchical layout is used to display a clear hierarchical relationship between entities. Users can adjust the display method of the hierarchical layout by setting parameters.

[0052] See Figure 3 As shown, the matrix layout is used to represent the relationships between large-scale entities.

[0053] Meanwhile, the interactive multi-dimensional visualization module of the domain knowledge model provides a GIS geographic view. For example... Figure 4 As shown, the GIS geographic view can display entities with latitude and longitude data. The left sidebar of the GIS geographic view is a tree view, which is used to display the conceptual model, attribute model, label model and relation model of the domain knowledge model.

[0054] When faced with massive amounts of data, data characteristics and relationships cannot be fully represented by graphs alone. For example... Figure 5 As shown, the domain knowledge model provides a hierarchical visualization function that is business logic oriented. This function can help users explore business information more accurately step by step and assist users in making decisions, helping users to clarify business logic in complex data and trace the origin of entity relationships.

[0055] To address diverse needs, domain knowledge models offer various perspectives for exploring data. When considering relationships between data, domain knowledge models clearly represent these relationships using graph structures, such as... Figure 6 As shown. When faced with exploring detailed data features, the domain knowledge model provides a table structure to present the data in detail, such as... Figure 7 As shown, the domain knowledge model provides a visualization method that combines diagrams and graphs. It uses a structural diagram between nodes to show the relationships between nodes, and tables to display detailed attribute information of nodes and relationships.

[0056] When faced with massive amounts of data, users of domain knowledge models sometimes do not need to observe all data of the same type simultaneously. For example... Figure 8 and Figure 9 As shown, the domain knowledge model provides structural views at both micro and macro scales. Users can view a micro view consisting of all nodes, or a macro view obtained by clustering a subset of nodes. The domain knowledge model offers various clustering methods, such as clustering similar nodes, related nodes, and multi-select nodes, allowing users to view views from multiple perspectives.

[0057] Multiple views and layouts provide users with different ways and perspectives to view the domain knowledge model, enabling them to gain a richer understanding of it. Multi-scale, multi-view knowledge visualization technology combines the extraction of different data types with the relevant features of the front-end framework to display data from multiple dimensions.

[0058] 2. Domain Knowledge Model Interactive Knowledge Reasoning and Exploration Module

[0059] Traditional knowledge graphs often lack precise data screening, which hinders user exploration and decision-making when faced with massive datasets. To address this issue, the interactive knowledge reasoning and exploration module of the domain knowledge model described in this invention provides multiple methods for accurately locating data information, including global filtering, local filtering, and ontology navigation. By storing the conceptual model in the cloud, users download portions of the model to explore the data. The system achieves a balance between efficiency and performance by classifying and storing relationships, entities, and models. The interactive knowledge reasoning and exploration module combines document-based and graph databases, enabling data filtering and loading at different granularities. This provides good interactivity while reducing redundant information and conserving machine resources.

[0060] Figure 10This paper illustrates the global filtering view described in this invention. The global filtering method uses an ontology model as a basis to perform precise conditional retrieval of data across the entire knowledge graph. Users can use this method to retrieve globally matching data. The domain knowledge model uses Janusgraph to set fine-grained retrieval conditions for the model's relationships. After parsing the retrieval conditions, the cloud system performs complex conditional queries on the relevant data in the document-based database. This approach enables complex queries on the domain knowledge model's data while simultaneously improving retrieval speed.

[0061] See Figure 13 As shown, the process of using a global filtering method for knowledge graph querying is as follows:

[0062] S11, Global Selection Concept;

[0063] S12. Select the concept-related relationships in the knowledge graph ontology model;

[0064] S13. Set filter conditions for attributes of concepts and relationships;

[0065] S14. Filter concepts and relationships in a document-based database to find entities and edges that meet the filtering criteria.

[0066] S15. In the graph database, select the triples (entity-edge-entity) that meet the requirements by type and graph structure. Combine the entities and edges that meet the requirements found in step S14 to select the data that meets the requirements.

[0067] In the domain knowledge model, in addition to global filtering without context, local filtering with starting entities is also provided. Figure 11 This illustration shows a partial filtering view as described in this invention. Partial filtering uses entities as a starting point to perform precise conditional indexing on data associated with those entities. Users can use partial filtering to "follow the clues" and uncover hidden facts and knowledge related to entity data. Partial filtering involves setting a starting entity, querying the relationship type and specific attributes of the starting entity in Janusgraph, progressively exploring the entity by setting conditions, and returning the results of the combined conditions found in the document-based database to the user.

[0068] See Figure 14 As shown, the process of using local filtering to query a knowledge graph is as follows:

[0069] S21. Select an entity in the workspace;

[0070] S22. Select local filtering, and then select the relationship related to the concept to which the entity belongs in the knowledge graph ontology model;

[0071] S23. Set filtering conditions for attributes of concepts and relationships;

[0072] S24. Filter concepts and relationships in a document-based database to find entities and edges that meet the filtering criteria.

[0073] S25. In the graph database, select the triples (entity-edge-entity) that meet the requirements by type and graph structure. Combine the entities and edges that meet the requirements found in step S24 to select the data that meets the requirements.

[0074] Domain knowledge models also provide ontology navigation functionality for rapid, coarse-grained data exploration. Figure 12 The ontology navigation view described in this invention is shown. The ontology navigation function performs coarse-grained data retrieval and filtering starting from entity nodes. Users can quickly load entity-related relationships and node data through ontology navigation.

[0075] See Figure 15 As shown, the process of using ontology navigation for knowledge graph querying is as follows:

[0076] S31. First, select the entity;

[0077] S32. Select the ontology navigation and choose the relationship to expand from the entity's related relationships.

[0078] The knowledge graph query system and method based on graph databases and document databases provided by this invention, through the use of interactive multidimensional visualization technology and interactive knowledge reasoning and exploration technology of domain knowledge models, can better display and query massive amounts of data when faced with knowledge graphs containing massive amounts of data. This enables users to efficiently and accurately find the information they need, clarify data relationships, assist users in decision-making, reduce the time users spend querying and processing data, and maximize the usability of massive data knowledge graphs.

[0079] The above embodiments are merely illustrative examples of the present invention. The present invention may also be implemented in other specific ways or forms without departing from its spirit or essential characteristics. Therefore, the described embodiments should be considered illustrative rather than limiting in any respect. The scope of the present invention should be defined by the appended claims, and any variations equivalent to the intent and scope of the claims should also be included within the scope of the present invention.

Claims

1. A knowledge graph query system based on graph databases and document databases with graph-based linkage, characterized in that, The system includes an interactive multidimensional visualization module for domain knowledge models and an interactive knowledge reasoning and exploration module for domain knowledge models. The interactive multidimensional visualization module of the domain knowledge model is used to query the structural relationships between entities and provides multiple ways to view the domain knowledge model, including force-guided layout view, hierarchical layout view and matrix layout view; The interactive multidimensional visualization module of the domain knowledge model also provides multiple perspectives for viewing the domain knowledge model, including GIS geographic view, hierarchical progressive view, graph structure view, table structure view, micro-scale view, and macro-scale view; wherein, the micro-scale view is used for users to view all nodes; the macro-scale view is used for users to view clustered nodes, and the clustering methods include clustering of nodes of the same type, clustering of related nodes, and clustering of multiple selected nodes; The interactive knowledge reasoning and exploration module of the domain knowledge model is used to accurately locate data information. The data exploration methods provided include global filtering, local filtering, and ontology navigation. The global filtering method is used to retrieve globally matching data, using the ontology model as a basis for precise conditional retrieval of data across the entire knowledge graph. Specifically, it includes the following steps: S11, Global Selection Concept; S12. Select the relationships related to the concept in the knowledge graph ontology model; S13. Set the filtering conditions for the attributes of the concept and the relationship; S14. Filter the concepts and relationships in the document database to filter out entities and edges that meet the filtering conditions; S15. In the graph database, select triples that meet the requirements by type and graph structure, and combine them with the entities and edges that meet the requirements queried in step S14 to select data that meets the requirements. The local filtering method uses entities as a starting point to perform precise conditional indexing on the data associated with those entities, specifically including the following steps: S21. Select an entity in the workspace; S22. Select local filtering, and then select the relationship related to the concept to which the entity belongs in the knowledge graph ontology model; S23. Set the filtering conditions for the attributes of the concept and the relationship; S24. Filter the concepts and relationships in the document database to select entities and edges that meet the filtering conditions; S25. In the graph database, select triples that meet the requirements by type and graph structure, and combine them with the entities and edges that meet the requirements queried in step S24 to select data that meets the requirements. The ontology navigation method performs coarse-grained data retrieval and filtering starting from entity nodes, specifically including the following steps: S31. Select entity; S32. Select the relationships to be expanded from the relationships related to the entity.

2. The knowledge graph query system based on graph database and document database linkage according to claim 1, characterized in that, The force-guided layout view is used to display ordinary physical structural relationships, the hierarchical layout view is used to display clear hierarchical relationships between entities, and the matrix layout view is used to display relationships between large-scale entities.

3. The knowledge graph query system based on graph database and document database linkage according to claim 2, characterized in that, The GIS geographic view is used to display entities with latitude and longitude data. The GIS geographic view includes a tree view to display the conceptual model, attribute model, label model, and relation model of the domain knowledge model.

4. The knowledge graph query system based on graph database and document database linkage according to claim 3, characterized in that, The hierarchical view is business logic oriented and is used to trace the origins of entity relationships.

5. The knowledge graph query system based on graph database and document database linkage according to claim 4, characterized in that, The graph structure view is used to display the relationships between entities; the table structure view is used to display detailed attribute information of entities and relationships.

Images