Tuesday, 25 September, 2012 - 15:00

Learning under Bias in NLP

Abstract: In NLP we rely on manually annotated data, e.g. treebanks. Such data is hard to come by, explaining recent interests in semi-supervised NLP. However, our labeled data is also (almost always) extremely biased. This talk presents bias correction techniques and discusses their applicability in NLP.


Anders Søgaard did his Ph.D. in 2007 at the University of
Copenhagen in mathematical linguistics. He has been a Senior
Researcher at the University of Potsdam and now works as an
Associate Professor at the University of Copenhagen. He was
recently awarded an European Research Council Starting