Investigating Distributed Approaches to Efficiently Extract Textual Evidences for Biomedical Ontologies

Abstract

Heterogeneous data resources in biomedicine become available both in structured and unstructured formats, such as scientific publications, healthcare guidelines, controlled vocabularies, and formal ontologies. Bridging the gaps among these heterogeneous data is useful to discovery implicit knowledge. To make this happen, efficient computational approaches are a necessity for applications in such a knowledge-and dataintensive domain. In this paper, we first define a particular task, relation alignment, which is to identify textual evidences for biomedical ontologies. Then, we investigate two parallel approaches for this task over distributed systems and present the details of their implementations. Moreover, we characterize the performance of our methods through extensive experiments, thereby allowing researchers to make a more informed choice in the presence of large-scale biomedical data.