Learning under Bias in NLP

Speaker:
Anders Søgaard
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.
Length:
01:35:58
Date:
25/09/2012
views: 3831

Images:
Preview of img-040.jpg
Image img-040.jpg
Preview of img-059.jpg
Image img-059.jpg
Attachments: (video, slides, etc.)
43,9M
760 downloads
188M
755 downloads
944M
825 downloads
851K
1162 downloads