Missouri Botanical Garden Open Conference Systems, TDWG 2013 ANNUAL CONFERENCE

Font Size: 
Implementing a semantic mediator to integrate flower-visiting records from South African natural history museums
Willem Coetzer, Deshendran Moodley, Aurona Gerber

Building: Grand Hotel Mediterraneo
Room: Sala dei Continenti
Date: 2013-11-01 09:39 AM – 09:52 AM
Last modified: 2013-10-07

Abstract


Specimen-records from different natural history museums take different interpretations of certain concepts, e.g. ‘pollinator’. This poses challenges for systems interoperability and automatic integration and analysis of these datasets . We created an ontology of concepts related to flower-visiting and pollen-bearing, and used this in a pilot implementation of a semantic mediation system for three specimen databases containing flower-visiting records. 

We created or used three ontologies in our semantic mediator, namely a Flower-Visiting and Pollen-Bearing domain ontology (fv:) , an Occurrence domain ontology and a separate Flower-Visiting Mapping ontology (fvm:). The Occurrence ontology was based on the Darwin Semantic Web Ontology and was used to represent the standard Darwin Core classes (Occurrence, IndividualOrganism, Taxon). The classes in the application ontology created a mapping mechanism, using the dwc:behavior property, to manually bridge the conceptual gap between the data in a given specimen database and the classes in the fv domain ontology. This means that each database had a different, database-specific mapping class (e.g. fvm:BehaviorMappingAlbany for the Albany Museum database), the instances of which represented the unique text strings describing the behaviour of insect specimens in that database (e.g. “On plant”, “On flowers”). 

A Java implementation, using the Java JENA API, demonstrates how the ontologies are used to automatically translate specimen records into semantically enriched ontology instances. The specimen record is converted into an instance of dsw:Occurrence (e.g. fv:PlantVisitorOccurrence, fv:FlowerVisitorOccurrence) using an instance of the mapping class. The mapping instance is also used to create an instance of the appropriate subclass of fv:PlantAssociationEvent (e.g. fv:PlantVisit, fv:FlowerVisit) in the domain ontology. 

The reasoner classifies instances of the narrowly defined concepts read from the databases (e.g. “Feeding on nectar” means that there was an instance of the FlowerNectarIngestorOccurrence class) as instances of the subsuming classes (FlowerVisitorOccurrence and PlantVisitorOccurrence), thereby creating integrated lists of instances of the more general classes from all three databases. These semantically enriched instances can now be queried and analysed for flower-visiting and pollination research without the representational differences and nuances inherent in each specimen database. The pilot application shows the practical value of ontologies for integrating and analysing heterogeneous biodiversity data from multiple specimen databases.