It has been demonstrated that grouping multiple steps of a typical NLP pipeline into one, joint structured prediction task can provide accuracy benefits. I'm curious where people feel joint inference has made the biggest difference so far. Alternatively, what tasks do people think would benefit most from being done jointly?

asked Jul 13 '12 at 15:27

David%20Belanger's gravatar image

David Belanger
61244

Be the first one to answer this question!
toggle preview

powered by OSQA

User submitted content is under Creative Commons: Attribution - Share Alike; Other things copyright (C) 2010, MetaOptimize LLC.