Intelligent ontology alignment to improve semantic data integration across research domains in biomedicine

Abstract

Abstract Ontology alignment is a key component of semantic web interoperability, with a long history of applications in traditional data integration tasks that address the problem of semantic heterogeneity. Ontology alignment tools take two ontologiesas input and determine alignments as output i.e. a set of correspondences between semantically related units of these ontologies. These correspondences can then be used to merge ontologies, link data across knowledge domains, answer semantic queries, navigate through knowledge graphs, and many more. The process of determining alignments, called matching, therefore is a basic requirement for linking knowledge across scientific domains covered by one or more ontologies. A matching exists if entities from different ontologies are semantically equivalent. The goal of this study is to find all semantically equivalent entity pairs given a source ontology Os and a target ontology Ot, each consisting of a set of entities. We propose to use Inverse Document Frequency (IDF) and Jaccard distance to find candidate entities with high precision and less computational effort. Our goal to cross-link different research fields in the biomedical and clinical domains.