Matching Biomedical Ontologies: Construction of Matching Clues and Systematic Evaluation of Different Combinations of Matchers
Peng Wang,Yunyan Hu,Shaochen Bai,Shiyi Zou
DOI: https://doi.org/10.2196/28212
IF: 3.2
2021-08-19
JMIR Medical Informatics
Abstract:Background Ontology matching seeks to find semantic correspondences between ontologies. With an increasing number of biomedical ontologies being developed independently, matching these ontologies to solve the interoperability problem has become a critical task in biomedical applications. However, some challenges remain. First, extracting and constructing matching clues from biomedical ontologies is a nontrivial problem. Second, it is unknown whether there are dominant matchers while matching biomedical ontologies. Finally, ontology matching also suffers from computational complexity owing to the large-scale sizes of biomedical ontologies. Objective To investigate the effectiveness of matching clues and composite match approaches, this paper presents a spectrum of matchers with different combination strategies and empirically studies their influence on matching biomedical ontologies. Besides, extended reduction anchors are introduced to effectively decrease the time complexity while matching large biomedical ontologies. Methods In this paper, atomic and composite matching clues are first constructed in 4 dimensions: terminology, structure, external knowledge, and representation learning. Then, a spectrum of matchers based on a flexible combination of atomic clues are designed and utilized to comprehensively study the effectiveness. Besides, we carry out a systematic comparative evaluation of different combinations of matchers. Finally, extended reduction anchor is proposed to significantly alleviate the time complexity for matching large-scale biomedical ontologies. Results Experimental results show that considering distinguishable matching clues in biomedical ontologies leads to a substantial improvement in all available information. Besides, incorporating different types of matchers with reliability results in a marked improvement, which is comparative to the state-of-the-art methods. The dominant matchers achieve F1 measures of 0.9271, 0.8218, and 0.5 on Anatomy, FMA-NCI (Foundation Model of Anatomy-National Cancer Institute), and FMA-SNOMED data sets, respectively. Extended reduction anchor is able to solve the scalability problem of matching large biomedical ontologies. It achieves a significant reduction in time complexity with little loss of F1 measure at the same time, with a 0.21% decrease on the Anatomy data set and 0.84% decrease on the FMA-NCI data set, but with a 2.65% increase on the FMA-SNOMED data set. Conclusions This paper systematically analyzes and compares the effectiveness of different matching clues, matchers, and combination strategies. Multiple empirical studies demonstrate that distinguishing clues have significant implications for matching biomedical ontologies. In contrast to the matchers with single clue, those combining multiple clues exhibit more stable and accurate performance. In addition, our results provide evidence that the approach based on extended reduction anchors performs well for large ontology matching tasks, demonstrating an effective solution for the problem.
medical informatics