Building A Contrasting Taxa Extractor for Relation Identification from Assertions: BIOlogical Taxonomy & Ontology Phrase Extraction System

In this paper, we present the methods we used to extract bacteria and biotopes names and then to identify the relation between those entities while participating to the BioNLP’13 Bacteria and Biotopes Shared Task. We used machine-learning based approaches for this task, namely a CRF to extract bacteria and biotopes names and a simple matching algorithm to predict the relations. We achieved poor results: an SER of 0.66 in sub-task 1, and a 0.06 F-measure in both sub-tasks 2 and 3.