Pavel Dobrokhotov

SIB, Geneva

Curators of biological databases are faced with a challenging task of browsing through hundreds of abstracts every day in order to select articles that will be used for annotation. Bio-text mining techniques can be used in order to assist curators in this task.

I will present the work we did in the context of medical annotation of Swiss-Prot. A description of the document processing chain and the classifier retained will be followed by a demonstration of a prototype currently field-tested in the Swiss Institute of Bioinformatics (SIB), Geneva. 

I will then present a prospecting work we started on discovering and highlighting important terms for document classification.