Using NSGA-III for optimising biomedical ontology alignment

• Published in: CAAI Transactions on Intelligence Technology Vol. 4; no. 3; pp. 135 - 141
• Main Authors: Xue, Xingsi, Lu, Jiawei, Chen, Junfeng
• Format: Journal Article
• Language: English
• Published: Beijing The Institution of Engineering and Technology 01.09.2019; John Wiley & Sons, Inc; Wiley
• Subjects: Alignment; anatomy track; biomedical concept mapping; biomedical information systems; biomedical ontology alignment; C1180 Optimisation techniques; C6170K Knowledge engineering techniques; C7140 Medical administration; C7330 Biology and medical computing; Coma; Concept mapping; Genetic algorithms; heterogeneous biomedical concepts; Information systems; large-scale biomedical ontology matching problem; Linguistics; Matching; medical computing; medical information systems; nondominated sorting genetic algorithm-III-based biomedical ontology matching; NSGA-III; ontologies (artificial intelligence); Ontology; Ontology Alignment Evaluation Initiative; ontology partitioning technique; ontology segment-matching problems; Optimization; Preferences; Similarity; Similarity measures; Sorting algorithms; Special Issue: Advances in Bio-inspired Heuristics for Computing; Taxonomy; large-scale biomedical ontology matching problem; NSGA-III; medical information systems; genetic algorithms; biomedical concept mapping; similarity measures; biomedical ontology alignment; biomedical information systems; ontology partitioning technique; ontologies (artificial intelligence); Ontology Alignment Evaluation Initiative; anatomy track; medical computing; nondominated sorting genetic algorithm-III-based biomedical ontology matching; ontology segment-matching problems; heterogeneous biomedical concepts
• ISSN: 2468-2322; 2468-2322
• DOI: 10.1049/trit.2019.0014

To support semantic inter-operability between the biomedical information systems, it is necessary to determine the correspondences between the heterogeneous biomedical concepts, which is commonly known as biomedical ontology matching. Biomedical concepts are usually complex and ambiguous, which makes matching biomedical ontologies a challenge. Since none of the similarity measures can distinguish the heterogeneous biomedical concepts in any context independently, usually several similarity measures are applied together to determine the biomedical concepts mappings. However, the ignorance of the effects brought about by different biomedical concept mapping's preference on the similarity measures significantly reduces the alignment's quality. In this study, a non-dominated sorting genetic algorithm (NSGA)-III-based biomedical ontology matching technique is proposed to effectively match the biomedical ontologies, which first utilises an ontology partitioning technique to transform the large-scale biomedical ontology matching problem into several ontology segment-matching problems, and then uses NSGA-III to determine the optimal alignment without tuning the aggregating weights. The experiment is conducted on the anatomy track and large biomedic ontologies track which are provided by the Ontology Alignment Evaluation Initiative (OAEI), and the comparisons with OAEI's participants show the effectiveness of the authors' approach.