Protein name tagging in the immunological domain

The research described in this paper addresses the following question: How well do generic protein/gene name taggers perform when they are applied to full-text articles from the sub-domain of immunology (a subdomain with its own distinctive protein nomenclature)?
\nTo answer this question we have created a new corpus – ImmunoTome
\n– consisting of ten full-text immunological articles in which the names of proteins have been manually annotated. Our results show that a single tagger – ABNER trained on the BioCreAtivE corpus – performs significantly better than the other taggers we evaluated when applied to ImmunoTome. ImmunoTome is available from immunominer.cryst.bbk.ac.uk/tome.html.