BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Modelling selectional preferences in a lexical hierarchy - Diarmui
 d Ó Séaghdha\, University of Cambridge
DTSTART:20120601T110000Z
DTEND:20120601T113000Z
UID:TALK38356@talks.cam.ac.uk
CONTACT:Ekaterina Kochmar
DESCRIPTION:The talk will describe Bayesian selectional preference models 
 that\nincorporate knowledge from a lexical hierarchy such as WordNet. Insp
 ired\nby previous work on probabilistic modelling with WordNet\, these\nap
 proaches are based either on "cutting" the hierarchy at an appropriate\nle
 vel of generalisation or on a "walking" model that selects a path from\nth
 e root to a leaf. In an evaluation comparing against human\nplausibility j
 udgements\, we show that the models presented here\noutperform previously 
 proposed comparable WordNet-based models\, are\ncompetitive with state-of-
 the-art selectional preference models and are\nparticularly well-suited to
  estimating plausibility for items that were\nnot seen in training.
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR