A Case-Study of Ontology-Driven Semantic Mediation of Flower-Visiting Data from Heterogeneous Data-Stores in Three South African Natural History Collections

Abstract

Abstract. The domain complexity and structural-and semantic heterogeneity of biodiversity data, as well as idiosyncratic legacy data-creation processes, present significant integration and interoperability challenges. In this paper we describe a casestudy of ontology-driven semantic mediation using records of flower-visiting insects from three natural history collections in South Africa. We establish a conceptual domain model for flower-visiting, expressed in an OWL ontology, and use it to semantically enrich the three data-stores. We show how this enrichment allows for the creation of an integrated flower visiting data set. We discuss how this ontology captures both implicit and explicit knowledge, how it can be used to identify and analyze highlevel flower-visiting behaviour, and ultimately to construct flower-visiting and pollination networks.Keywords: biodiversity information, semantic mediation, ontology, plant-insect interactions, pollination IntroductionThe challenges of integrating, or making interoperable, distributed, heterogeneous sources of biodiversity-and ecological data have been described. Biodiversity is a complex domain and is no different from other domains in that users encode different definitions of the same concepts [3], which frustrates efforts to integrate data.We present a case study of three data-stores of flower-visiting insect specimens. All three data-stores consistently contained the names of the plant species, termed hostplants, with which both flower-visiting and non-flower-visiting insect specimens were associated. Whereas flower-visiting records were not explicit in most records of two data-stores, most records of the third data-store contained explicit, easily distinguishable flower-visiting data. To develop a semantic mediation solution, we created the first version of an OWL ontology containing concepts related to flower-visiting and the utilization of flower products, as well as the bearing of pollen by insect vectors. Our work will facilitate the construction of a system to bring about interoperability between distributed and heterogeneous biodiversity data-stores and systems. This will enable biodiversity scientists to more easily extract and analyze the behaviour of flower-visiting insects. Such a system would allow flower-visiting and pollination networks to be automatically assembled and compared.Outline. In Section 2 we sketch the background against which the need for our study emerged, discuss previous work in biodiversity semantics, and introduce our casestudy of interoperability of flower-visiting data. Section 3 begins by describing the domain of flower-visiting and pollination, including our scope, before explaining the process of ontology construction. Expert-and implicit knowledge is highlighted. The usefulness of the concepts in the ontology is discussed in Section 4, by linking data from the data-stores to classes in the ontology. Finally we discuss our approach to a potential solution, including areas where future work is required, and conclude. Background 2