BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Bayesian non-parametric models for parsing and translation - Trevo
 r Cohn\, University of Sheffield
DTSTART:20091023T110000Z
DTEND:20091023T120000Z
UID:TALK20470@talks.cam.ac.uk
CONTACT:Laura Rimell
DESCRIPTION:Context free grammars have long been popular for modelling nat
 ural language syntax and translation between human lanuages. However\, the
  underlying independencies assumed by the model are much too stringent for
  accurate data modelling. Considerable research effort has focussed on usi
 ng linguistic intuitions to enrich CFGs\, resulting in state-of-the-art pa
 rsing performance. In this talk\, I take a different approach by learning 
 an enriched grammar directly from the data without result to linguistic kn
 owledge. Instead the grammar is an emergent structure\, found by unsupervi
 sed inference in a Bayesian model of tree-substitution grammar (TSG\; a.k.
 a. DOP). Bayesian methods provide an elegant and theoretically principled 
 way to model TSG by including a prior over the grammar and integrating ove
 r uncertain events. In this talk I'll describe non-parametric Bayesian mod
 els for two related tasks: 1) learning a TSG for syntactic parsing and 2) 
 learning a synchronous TSG for machine translation. The models learn compa
 ct and simple grammars\, uncovering latent linguistic structures and in do
 ing so outperform competitive baselines. \n\nThis is joint work with Phil 
 Blunsom and Sharon Goldwater. \n\nhttp://www.dcs.shef.ac.uk/~tcohn/
LOCATION:SW01\, Computer Laboratory
END:VEVENT
END:VCALENDAR