reminder tomorrow at noon<br><br>On Monday, March 31, 2014, Dani Yogatama <<a href="mailto:dyogatama@cs.cmu.edu">dyogatama@cs.cmu.edu</a>> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div dir="ltr"><div style="font-family:arial,sans-serif;font-size:13px"><b>CL+NLP Lunch </b>(<u style="color:rgb(17,85,204)"><a href="http://www.cs.cmu.edu/~nlp-lunch/" target="_blank">http://www.cs.cmu.edu/~<span style="background-color:rgb(255,255,204);color:rgb(34,34,34)">nlp</span>-<span style="background-color:rgb(255,255,204);color:rgb(34,34,34)">lunch</span>/</a></u>)</div>
<div style="font-family:arial,sans-serif;font-size:13px"><b>Speaker</b><font face="arial, sans-serif">: </font>Dan Garrette, The University of Texas at Austin<br><b>Date</b><font face="arial, sans-serif">: Thursday, April 10, 2013</font><br>
<b>Time</b><font face="arial, sans-serif">: 12:00 noon</font><br><b>Venue</b><font face="arial, sans-serif">: GHC 6501</font><br></div><span style="font-family:arial,sans-serif;font-size:13px"><div><br></div><b>Title</b>: Learning Combinatory Categorial Grammars from Wrak Supervision</span><div style="font-family:arial,sans-serif;font-size:13px">
<br><b>Abstract</b>: <div>Grammar learning is a well-studied problem in NLP, but the task is<br>particularly difficult for low-resource languages. In this talk, I<br>will discuss our current work in learning combinatory categorial<br>
grammars from various forms of weak supervision. First, I will show<br>how we can learn good sequence-based CCG supertaggers by encoding<br>universal, inherent properties of the CCG formalism as priors over<br>both the appearance of supertags and the transitions between<br>
supertags. These universal priors can, in turn, be combined with<br>corpus-specific knowledge derived from available (partial) tag<br>dictionaries and unannotated text to further improve tagging<br>performance. Then, I will discuss our current efforts to extend these<br>
principles to tree grammars to learn CCG parsers. Finally, I will<br>discuss how simple annotations --- particularly those given in the<br>Graph Fragment Language developed at CMU --- may be used to help learn parsers under extremely tight annotation budgets.<br>
<br>This work is in collaboration with Jason Baldridge, Chris Dyer, and Noah Smith.<br><br><b>Biography</b>: </div><div>Dan is a a Computer Science Ph.D. student at The University of Texas<br>at Austin. His research focuses on Natural Language Processing and<br>
Machine Learning. He was a Best Talk Award nominee at NAACL this past year. He thinks slides should have less text and more animations.<br></div><div><br></div></div></div>
</blockquote>