Knowledge Graph Vector Representation Ontology Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine learning methods using knowledge graphs struggle with achieving high accuracy in vector representation, particularly in unorganized graphs, as they fail to accurately handle abstraction levels of entities and relations, leading to decreased training accuracy and inability to distinguish between entities with different abstraction levels.
Innovation Solution
The introduction of ontology-based constraints, specifically through entailment determination and class hierarchy usage, to improve the accuracy of vector calculations by emphasizing more specific relations and entities, ensuring that the difference in vector representations aligns with the specified constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional machine learning methods are used for knowledge graph embedding, then the processing speed is fast, but the manufacturing precision of vector representation is low
Solution Approach 1:
The patent applies preliminary action by pre-determining entailment relationships between relations and class hierarchies before the main vector generation process. The determination unit pre-identifies which relations entail other relations and establishes class hierarchy structures in advance, so that during machine learning, these pre-established constraints can be directly applied to guide vector updates, improving representation accuracy without adding significant computational complexity during training.
Solution Approach 2:
The patent introduces an intermediary determination unit that acts as a mediator between the knowledge graph structure and the machine learning process. This determination unit analyzes entailment relationships and class hierarchies, then translates these structural relationships into constraints that guide the vector update process, effectively bridging the gap between knowledge graph ontology and embedding learning.
2Manufacturing precision
If ontology-based constraints are introduced to improve vector calculation accuracy, then the manufacturing precision increases, but the device complexity increases
Solution Approach 1:
The patent applies local quality by applying different types of constraints to different parts of the knowledge graph based on their specific characteristics. The determination unit identifies which relations have entailment relationships and which entities belong to hierarchical class structures, then applies appropriate constraints locally to each triple during vector updates, rather than applying uniform constraints across the entire knowledge graph.
3Measurement precision
If conventional embedding methods are used, then the device complexity is low, but the measurement precision of entity and relation relationships is low
Solution Approach 1:
The patent implements feedback by using the determined entailment relationships and class hierarchy information to continuously guide and adjust the vector update process. During machine learning, the system references the pre-determined constraints to ensure that vector updates maintain consistency with the knowledge graph's ontological structure, providing continuous feedback that improves relationship prediction accuracy.
Data Source
AI summary
An apparatus determines which of the first triple and the second triple is associated with more specific information based on a first comparison between a first relation between first two entities included in the first triple and a second relation between second two entities included in the second triple according to an occurrence status of each of relations between entities in a specific set of classes included in the knowledge graph and a second comparison between a first entity connected to any one of the first two entities and a second entity connected to any one of the second two entities, and generates vectors representing elements of the first triple and vectors representing elements of the second triple by machine learning based on a constraint that a difference in the vectors of the first triple is smaller than a difference in the vectors of the second triple.