Ontology alignment for knowledge representation and integration : applications to biomedical text

• Main Author: Chua, Watson Wei Khong.
• Other Authors: School of Computer Engineering
• Format: Theses and Dissertations
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering::Computer science and engineering::Information systems::Information systems applications
• Online Access: http://hdl.handle.net/10356/52911
• Institution: Nanyang Technological University

Ontology alignment refers to the task of finding correspondences between entities in different ontologies. It facilitates inter-operability between applications using different ontologies as the correspondences allow them to understand one another’s data. In this dissertation, we focus on the discovery of both equivalence and subsumption correspondences between the concepts of different ontologies and use them for knowledge representation and integration in the biomedical domain. For equivalence correspondence discovery, we introduce a novel technique called Bridging Ontologies Alignment Technique (BOAT) which is state-of-the-art in terms of accuracy and speed compared to existing techniques. Existing approaches for equivalence correspondence discovery consider a pair of concepts to be equivalent if their names are identical or share many common words. Their accuracy can be improved by taking the structures of the ontologies into account. In BOAT, we do so by combining the word-based comparison with a structural comparison to determine the equivalence between concepts. Given a pair of concepts, we collect difference words that are found in the name of one concept but not in the other, and determine if each of the words distinguishes the concepts, by using the ontology structures. Concept pairs with no such distinguishing difference words are then considered equivalent. This enhancement helped BOAT achieve the highest F-Measure of 0.88 among the top matchers in the anatomy track of the OAEI competition. BOAT is also one of the fastest matchers as it reduces the time taken for matching large ontologies using a novel candidate selection technique. This technique selects only concept pairs with high similarities for comparison, based on the Vector Space Model.