BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Leveraging a Semantically Annotated Corpus to Disambiguate Preposi
 tional Phrase Attachment - Guy Emerson\, University of Cambridge
DTSTART:20150409T110000Z
DTEND:20150409T113000Z
UID:TALK58549@talks.cam.ac.uk
CONTACT:Tamara Polajnar
DESCRIPTION:Accurate parse ranking requires semantic information\, since a
  sentence may have many candidate parses involving common syntactic constr
 uctions. In this paper\, we propose a probabilistic framework for incorpor
 ating distributional semantic information into a maximum entropy parser. F
 urthermore\, to better deal with sparse data\, we use a modified version o
 f Latent Dirichlet Allocation to smooth the probability estimates. This LD
 A model generates pairs of lemmas\, representing the two arguments of a se
 mantic relation\, and can be trained\, in an unsupervised manner\, on a co
 rpus annotated with semantic dependencies. To evaluate our framework in is
 olation from the rest of a parser\, we consider the special case of prepos
 itional phrase attachment ambiguity. The results show that our semanticall
 y-motivated feature is effective in this case\, and moreover\, the LDA smo
 othing both produces semantically interpretable topics\, and also improves
  performance over raw co-occurrence frequencies\, demonstrating that it ca
 n successfully generalise patterns in the training data.\n\n
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR