A method for densifying a geographic knowledge graph based on classification knowledge

Abstract

The present application belongs to the field of artificial intelligence and geographic information technology, and specifically relates to a geographic knowledge graph densification method based on classification knowledge. The method comprises: constructing an element set of classification knowledge; calculating and generating a geographic knowledge graph semantic enhancement node; generating a geographic knowledge graph semantic enhancement relationship; and fusing to generate a dense geographic knowledge graph. The present application provides a scientific, universal and efficient knowledge graph optimization method, which can improve the density and integrity of the geographic knowledge graph, and provide data support for downstream applications such as knowledge graph driven geographic information analysis, spatial reasoning and recommendation system.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence, specifically relating to a semantically dense knowledge graph completion method. Background Technology

[0002] With the widespread application of knowledge graphs across various fields, their density has become a key factor in improving their reasoning and application capabilities. Knowledge graphs represent entities and their relationships in a structured manner, helping machines better understand and utilize information. However, in many fields, knowledge graphs currently suffer from sparsity, meaning that entity and relational data are relatively scarce, leading to lower accuracy in reasoning results and lower efficiency in knowledge transfer. This sparsity problem is particularly prominent in complex, multi-domain integrated application scenarios.

[0003] To improve the semantic density of knowledge graphs, researchers have increasingly recognized that introducing finer-grained classifications and semantic features is an effective approach. A classification index system is a framework that can refine the classification of entities and relationships in a knowledge graph based on different semantic levels. Through classification index systems, multi-level and multi-dimensional semantic information in knowledge graphs can be captured, making the relationships within the graph more cohesive and increasing its density.

[0004] However, most existing knowledge graph completion methods rely solely on existing structural information, neglecting the introduction of domain-specific classification semantic features. Furthermore, current mainstream methods are often limited to a single data source, failing to effectively utilize the rich information in multimodal data (such as text and images) to further enhance the density of the graph. Due to the complexity of knowledge systems across different domains, single-source indicator systems cannot meet the needs of multidimensional knowledge representation.

[0005] Therefore, there is an urgent need for a general method for knowledge graph densification based on classification index systems. This method should not only systematically collect classification index systems from national standards, academic papers, and multimodal data, but also apply these index systems to knowledge graphs through a reasonable classification framework, thereby significantly improving the density of the graph. In view of the shortcomings of the above-mentioned technologies, those skilled in the art are dedicated to developing a general, scientific, and automatic method for recommending ecological civilization models. Summary of the Invention

[0006] This invention discloses a knowledge graph densification method based on a knowledge graph and classification index framework, aiming to solve the sparsity problem of existing methods when processing complex knowledge systems. Current knowledge graphs are prone to relationship sparsity issues in large-scale data processing, affecting the performance of reasoning, querying, and knowledge discovery. To improve the density of knowledge graphs, this invention proposes a knowledge graph densification method based on an authoritative classification index framework, which can significantly improve the connectivity and semantic richness between nodes in the graph.

[0007] To achieve the above objectives, this invention provides a semantically dense knowledge graph completion method, comprising the following steps:

[0008] S1. Construct a set of elements for classification knowledge;

[0009] S2. Calculate and generate semantically enhanced nodes for the geographic knowledge graph;

[0010] S3, the relationship between semantic enhancement in generating geographic knowledge graphs;

[0011] S4. Integrate and generate a dense geographic knowledge graph.

[0012] Furthermore, the specific method of step S1 includes the following sub-steps:

[0013] S11. Obtaining Classification Knowledge Elements To form a collection of categorized knowledge The details are as follows:

[0014]

[0015]

[0016] in, Let i be the set of categorized knowledge; i is the index number of the categorized knowledge in the set of categorized knowledge. This represents the i-th category knowledge in the category knowledge set; It is the number of categorized knowledge items in the categorized knowledge set; The denot represents the semantics corresponding to the i-th category knowledge; j is the index number of the value range interval of the category knowledge. Let J represent the minimum and maximum values ​​of the range of the j-th category semantics of the i-th category knowledge in the category knowledge set. Indicates the range of values; "Indicates that the boundary is uncertain whether it is closed; express The name corresponding to the range of values; Indicates the number of intervals for the attribute;

[0017] S12. Import the geographical knowledge graph, as detailed below:

[0018]

[0019] in, A set of triples for geographical knowledge graphs; For the index number of the triple in the geographical knowledge graph; This is the g-th triplet in GeoKG; These represent the g-th geographic entity, its relation, and its attribute value in GeoKG. This refers to the number of all triples in GeoKG;

[0020] S13. Determine the set E of geographic entities to be completed and the geographic entities to be completed from GeoKG. The associated triples The details are as follows:

[0021]

[0022]

[0023] Where E is the set of geographic entities to be completed; CT is the set of triples associated with the geographic entities to be completed; and k is the index number of the geographic entities to be completed in GeoKG. Let k be the k-th geographic entity in GeoKG; The relation of the k-th geographic entity in GeoKG; This represents the attribute value of the k-th geographic entity in GeoKG. This represents the number of geographic entities to be completed in GeoKG. This represents the set of index numbers for geographic entities to be completed in GeoKG;

[0024] S14. Filter the valid sets of classification knowledge. The specific filtering rules are as follows:

[0025]

[0026] in, It is a set of effective classified knowledge selected from a set of classified knowledge.

[0027] S15. Establish a connection between CT and S through semantic content. index:

[0028]

[0029] in, It is a mapping established through semantic content between the triples associated with the geographic entities to be completed and the set of classification knowledge.

[0030] 2. Further, the specific method of step S2 includes the following sub-steps:

[0031] S21. Determine the classification knowledge in the valid classification knowledge set. ,enter Generate semantically enhanced nodes; the specific generation function is as follows:

[0032]

[0033] in, These are semantically enhanced nodes in the geographic knowledge graph generated by E and S; This indicates an empty set, and no nodes are generated.

[0034] 3. Further, the specific method of step S3 includes the following sub-steps:

[0035] S31. Determine the classification knowledge in the valid classification knowledge set. ,enter Generate semantically enhanced relationships, that is, the edges between the geographic entities to be completed and the semantically enhanced nodes. The specific generation function is as follows:

[0036]

[0037] in, It is a semantically enhanced relation in the geographic knowledge graph generated by E and S; The direction of semantic enhancement relation, the direction is determined by point to .

[0038] 4. Further, the specific method of step S4 includes the following sub-steps:

[0039] S41. Combine the semantically enhanced nodes generated in S21 with those in GeoKG. Merge the nodes to generate the complete set of nodes in the geographic knowledge graph, as follows:

[0040]

[0041] in, To complete the final set of nodes in the geographic knowledge graph, It is a set of semantically enhanced nodes in a geographic knowledge graph generated by E and S. A collection of geographic entities in GeoKG;

[0042] S42. Combine the semantically enhanced relations generated in S31 with those in GeoKG. Merge the results to generate the complete set of relationships in the geographic knowledge graph, as follows:

[0043]

[0044] in, To complete the overall set of relationships in the geographical knowledge graph, It is a set of semantically enhanced relations in a geographic knowledge graph generated by E and S. This is a set of relations in GeoKG.

[0045] Beneficial Effects: The geographical knowledge graph densification method based on classification knowledge provided by this invention has the following beneficial effects compared with existing methods:

[0046] Enhancing the semantic density of knowledge graphs: Through the classification semantic classification and automatic expansion mechanism proposed in this invention, the blank areas in knowledge graphs can be effectively filled, especially in fields with high graph sparsity, achieving higher semantic coverage and relational connectivity, and enhancing the structural integrity of knowledge graphs.

[0047] (1) Precise semantic classification: This method constructs a classification system corpus and uses classification algorithms to automatically classify knowledge, ensuring the accurate classification and organization of semantic elements. At the same time, by verifying and adjusting the classification results, the classification accuracy is improved, ensuring the rigor and usability of the knowledge graph structure.

[0048] (2) Multi-source data integration and strong knowledge expansion capability: This method can automatically obtain classification indicators and related knowledge from multiple heterogeneous data sources, integrate data from different sources, increase the breadth and depth of the knowledge graph, and enhance its ability in multi-domain and multi-dimensional knowledge representation.

[0049] (3) Enhancing Reasoning Ability by Complexifying Graph Relationships: This method further enriches the ways in which entities in the graph are associated by generating multi-level complex relationships. Through the application of the reasoning engine and multi-level relationship generation algorithm, new knowledge triples can be dynamically generated, enhancing the reasoning ability and scalability of the knowledge graph.

[0050] (4) Visualization and densification effect verification: Through the visualization of the graph structure and density analysis tools, this method can intuitively present the optimization effect of the knowledge graph. Implementers can monitor the densification process of the graph in real time and conduct accurate effect evaluation to ensure the continuous improvement of the knowledge graph. Attached Figure Description

[0051] Figure 1 A flowchart of the overall process for semantic densification methods for knowledge graphs.

[0052] Figure 2 Comparison chart of semantic density enhancement performance of geographical knowledge graphs Detailed Implementation

[0053] The invention will now be further described with reference to the accompanying drawings.

[0054] See Figures 1-2 This invention provides a method for densifying geographic knowledge graphs based on classification knowledge, which can be broadly divided into three aspects: constructing a set of semantically enhanced elements based on classification knowledge, calculating and generating semantically enhanced nodes for the geographic knowledge graph, and determining the semantically enhanced relationships for the generated geographic knowledge graph. Specifically, taking a partial ecological civilization model knowledge graph (ECMKG) as an implementation example, the method includes the following steps:

[0055] S1. Constructing a set of elements for classification knowledge.

[0056] S11. Various classification knowledge was obtained through manual searching from national standards, industry standards, thematic maps, high-quality papers, official reports, and news reports, and the classification formula was summarized as: The specific knowledge regarding population and city size classification is as follows: .in, = 1 indicates that there is a category of knowledge; = 2, indicating that the first category of knowledge has two categories;

[0057] S12. Import the Ecological Civilization Model Knowledge Graph ECMKG (ECMKG is a graph structure composed of a series of nodes and relationships);

[0058] S13. Determine the set of geographic knowledge graph objects E to be completed and its corresponding set of triples from ECMKG. }, specifically as Where EN=3, the set E of objects to be completed in the geographic knowledge graph is determined to be { };

[0059] S14. Determine the effective set of semantic enhancement elements, because... According to the rules By filtering classification knowledge, population and city classification rules can be incorporated into the semantic enhancement element set. ,therefore ;

[0060] S15, For each An index is created with S obtained from S14, that is, the population and city classification rules are associated with the triplet of the geographic entity to be completed, and an index is created through semantics.

[0061] S2. Calculate and generate semantically enhanced nodes for the geographic knowledge graph.

[0062] S21, Input and =[(City X, County Y, Number of beds (unit: beds / thousand people), 3.2),(City D, County E, Number of beds (unit: beds / thousand people), 8.4),(City A, County B, Number of beds (unit: beds / thousand people), 7.4)] (partial fields), substitute them into the semantic enhancement node model of the geographic knowledge graph, and the output is the complete set of node names. .

[0063] S3. Generate semantically enhanced relationships for geographic knowledge graphs.

[0064] S31, Input and as well as (Partial fields) are substituted into the semantically enhanced relation model for generating a geographic knowledge graph, and the output is the set of relations between each node. .

[0065] S4. Merge and generate a dense geographic knowledge graph.

[0066] S41. Based on the complete node name set N, use Neo4j to create new geographic knowledge graph nodes one by one, merge them with ECMKG, and generate the complete node set nodeSet;

[0067] S42. Complete the set of key-value pairs of the relation <name, direction> in G, find the corresponding geographic entities in Neo4j according to the direction, create new geographic knowledge graph relations one by one, merge them with ECMKG, and generate the complete ECMKG relation set relationSet.

[0068] S5. Evaluation of the effect of semantic densening of knowledge graphs

[0069] S51. To verify the effectiveness of this method, we select a graph global sparsity function from graph theory to demonstrate the method's effect on graph density. The graph global sparsity function is as follows:

[0070]

[0071] in, To determine global sparsity; This represents the number of edges in the graph. This represents the number of nodes in the graph.

[0072] S52. Count the number of nodes and edges in the original ECMKG and the completed ECMKG, and input them into the graph global sparsity function respectively. The global sparsity value of the original ECMKG is 8.3333%, and the global sparsity value of the enhanced dense ECMKG is 9.3750%, which verifies the effectiveness of this method. See details in [link to documentation]. Figure 2 .

[0073] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for densifying geographical knowledge graphs based on classification knowledge, comprising the following steps: S1. Construct a set of geographical knowledge graph elements for categorized knowledge, specifically including the following steps: S11. Obtaining Classification Knowledge Elements To form a collection of categorized knowledge The details are as follows: in, Let i be the set of categorized knowledge; i is the index number of the categorized knowledge in the set of categorized knowledge. This represents the i-th category knowledge in the category knowledge set; It is the number of categorized knowledge items in the categorized knowledge set; The denot represents the semantics corresponding to the i-th category knowledge; j is the index number of the value range interval of the category knowledge. Let J represent the minimum and maximum values ​​of the range of the j-th category semantics of the i-th category knowledge in the category knowledge set. Indicates the range of values; "Indicates that the boundary is uncertain whether it is closed; express The name corresponding to the range of values; Indicates the number of intervals for the attribute; S12. Import the geographical knowledge graph, as detailed below: in, A set of triples for geographical knowledge graphs; For the index number of the triple in the geographical knowledge graph; This is the g-th triplet in GeoKG; These represent the g-th geographic entity, its relation, and its attribute value in GeoKG. This refers to the number of all triples in GeoKG; S13. Determine the set E of geographic entities to be completed and the geographic entities to be completed from GeoKG. The associated triples The details are as follows: Where E is the set of geographic entities to be completed; CT is the set of triples associated with the geographic entities to be completed; and k is the index number of the geographic entities to be completed in GeoKG. Let k be the k-th geographic entity in GeoKG; The relation of the k-th geographic entity in GeoKG; This represents the attribute value of the k-th geographic entity in GeoKG. This represents the number of geographic entities to be completed in GeoKG. This represents the set of index numbers for geographic entities to be completed in GeoKG; S14. Filter the valid sets of classification knowledge. The specific filtering rules are as follows: in, It is a set of effective classified knowledge selected from a set of classified knowledge. S15. Establish a connection between CT and S through semantic content. index: in, It is a mapping established through semantic content between the triples associated with the geographic entities to be completed and the set of classification knowledge. S2. Calculate and generate semantically enhanced nodes for the geographic knowledge graph; S3, the relationship between semantic enhancement in generating geographic knowledge graphs; S4. Integrate and generate a dense geographic knowledge graph.

2. The method for densifying a geographic knowledge graph based on classification knowledge according to claim 1, wherein step S2 specifically includes the following steps: S21. Determine the classification knowledge in the valid classification knowledge set. ,enter Generate semantically enhanced nodes; the specific generation function is as follows: in, These are semantically enhanced nodes in the geographic knowledge graph generated by E and S; This indicates an empty set, and no nodes are generated.

3. The method for densifying geographical knowledge graphs based on classification knowledge according to claim 2, characterized in that, Step S3 specifically includes the following steps: S31. Determine the classification knowledge in the valid classification knowledge set. ,enter Generate semantically enhanced relationships, that is, the edges between the geographic entities to be completed and the semantically enhanced nodes. The specific generation function is as follows: in, It is a semantically enhanced relation in the geographic knowledge graph generated by E and S; The direction of semantic enhancement relation, the direction is determined by point to .

4. The method for densifying geographical knowledge graphs based on classification knowledge according to claim 3, characterized in that, Step S4 specifically includes the following steps: S41. Combine the semantically enhanced nodes generated in S21 with those in GeoKG. Merge the nodes to generate the complete set of nodes in the geographic knowledge graph, as follows: in, To complete the final set of nodes in the geographic knowledge graph, It is a set of semantically enhanced nodes in a geographic knowledge graph generated by E and S. A collection of geographic entities in GeoKG; S42. Combine the semantically enhanced relations generated in S31 with those in GeoKG. Merge the results to generate the complete set of relationships in the geographic knowledge graph, as follows: in, To complete the overall set of relationships in the geographical knowledge graph, It is a set of semantically enhanced relations in a geographic knowledge graph generated by E and S. This is a set of relations in GeoKG.

Images