A Novel Ontology Matching Model to Address Ontology Heterogeneity

초록

This study introduces a novel ontology matching model designed to address ontology heterogeneity by leveraging both textual and structural information within ontologies, alongside external data. The model employs a word embedding approach to refine word vectors for enhanced discrimination between semantically similar and associative descriptions. Additionally, it adopts BERT for generating dynamic word vectors, enabling the nuanced distinction of polysemous terms. Our model calculates structural similarity by transforming ontologies into graph structures and applying the SimRank algorithm to calculate the entities' structural similarity within these graphs. The matching process employs a stable matching algorithm to secure stable one-to-one correspondences, while one-to-many matches are determined through similarity thresholds and comparative analysis

키워드