Simon Walk, Postdoctoral Scholar
Biomedical Informatics Research
Stanford University

Abstract:
One of the key factors that defined the success of the semantic web is the efficient and easy (modular) reuse of existing structured knowledge bases. This is particularly important in the biomedical domain, where ontologies model highly specialized domain knowledge for a variety of different applications. Hence, finding and identifying ontologies, which are best suited for a given set of requirements, is a crucial task for biomedical researchers and practitioners.
In a foray to tackle this problem, the National Center for Biomedical Ontology has developed BioPortal—an online repository, designed to store, explore, maintain and discover biomedical ontologies. However, to be able to augment practitioners and researchers in their tasks of exploring and selecting ontologies for a given task, we first have to better understand how they interact with and make use of the functionality of currently available data repositories.
We address this problem by making use of Markov chains, random surfers, stationary distributions and clustering techniques to model, analyze and discuss different browsing behaviors of users on BioPortal.
In particular, the contributions of this work are three-fold: First, we present an in-depth empirical analysis of BioPortal. Second, we analyze and interpret the browsing behavior of users for different ontologies on BioPortal. Finally, we cluster and classify users according to their browsing behaviors and discuss the obtained results.