TX Task: Automatic detection of focus organisms in biomedical publications

Abstract

In biomedical information extraction (IE), a central problem is the disambiguation of ambiguous names for domain specific entities, such as proteins, genes, etc. One important dimension of ambiguity is the organism to which the entities belong: in order to disambiguate an ambiguous entity name (e.g. a protein), it is often necessary to identify the specific organism to which it refers.
In this paper we present an approach to the detection and disambiguation of the focus organism(s), i.e. the organism(s) which are the subject of the research described in scientific papers, which can then be used for the disambiguation of other entities.
The results are evaluated against a gold standard derived from IntAct annotations. The evaluation suggests that the results may already be useful within a curation environment
and are certainly a baseline for more complex approaches.