Ontology Matching via Multi-Score Entity Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ontology matching systems focus primarily on computing equivalence and subsumption relationships, often missing entities related by other relationships, leading to inefficiencies in semantically linking large real-world ontologies from heterogeneous sources.
Innovation Solution
A computer-implemented method and apparatus for matching ontologies by determining similarity vectors, confidence scores, and relation scores for each pair of entities across different ontologies, generating a mapping ontology based on these scores to establish relationships beyond traditional equivalence and subsumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing ontology matching systems focus on computing equivalence and subsumption relationships, then the matching process is simplified and faster, but the system misses entities that are related by other relationships (e.g., part-of, subsumed-by, or other semantic connections)
Solution Approach 1:
The patent segments the relationship detection process into multiple independent similarity computation modules, each handling a specific type of relationship (equivalence, subsumption, part-of, etc.). This allows the system to compute all relationship types systematically without overwhelming complexity, improving both speed and completeness by dividing the matching task into manageable segments that can be processed in parallel
2Measurement precision
If the system computes similarity vectors and relation scores for all entity pairs across large ontologies, then the accuracy of ontology matching is improved, but the computational complexity and processing time increase significantly
Solution Approach 1:
The patent performs preliminary actions by pre-computing and storing similarity vectors for entity pairs during an initial processing phase. These pre-computed similarity vectors are then reused during the main matching process, eliminating the need to recompute similarities for all pairs and significantly reducing the computational complexity while maintaining high matching accuracy
Solution Approach 2:
The patent transforms the matching problem from a single high-dimensional similarity computation into multiple lower-dimensional steps: first computing similarity vectors in one dimension, then using confidence scores as a filtering dimension, and finally computing relation scores only for promising pairs. This dimensional decomposition reduces overall computational complexity while preserving accuracy
3Reliability
If the system uses multiple similarity metrics and confidence scores to evaluate entity pairs, then the reliability of matching results is improved, but the processing time and computational resources required increase
Solution Approach 1:
The patent applies partial action by computing multiple similarity metrics and confidence scores only for entity pairs that meet certain preliminary criteria (e.g., data type compatibility, name similarity thresholds). This selective computation approach ensures high reliability for the most promising matches while avoiding unnecessary computational resources spent on clearly unrelated pairs, thus reducing overall processing time
Data Source
AI summary
A computer implemented method of matching ontologies is disclosed. The method includes, for each pair of entities of a first ontology and a second ontology, wherein each pair of entities includes a first entity from a first plurality of entities of the first ontology and a second entity from a second plurality of entities of the second ontology, and wherein the first entity and the second entity of each pair of entities is of a same data type: (1) determining a vector of similarities for the pair of entities; (2) determining a confidence score for the vector of similarities; (3) determining a relation score for each relation type based on the vector of similarities to measure relatedness between the first entity of the pair of entities and the second entity of the pair of entities; and (4) generating a mapping ontology based on the relation type, the relation score, and the confidence score of each pair of entities.