BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Learning Commonsense Event Schemas from Unlabeled Text - Nate Cham
 bers\, US Naval Academy
DTSTART:20170120T120000Z
DTEND:20170120T130000Z
UID:TALK69898@talks.cam.ac.uk
CONTACT:Kris Cao
DESCRIPTION:The early years of research in natural language understanding 
 focused on big picture representations like scripts and frames to drive la
 nguage understanding. These formalisms motivated a rich line of research\,
  but they suffered from brittle hand-coded structure. Most recently\, seve
 ral lines of research have focused on new computational methods to learn s
 cript-like knowledge. Event schema induction\, for instance\, is the task 
 of learning high-level representations of complex events (e.g.\, a bombing
 ) and their entity roles (e.g.\, perpetrator and victim) from unlabeled te
 xt. As is usual\, many methods have been proposed\, and different event re
 presentations and learning approaches have been applied. This talk will mo
 tivate why such schema learning is important to NLU\, and present a partic
 ular generative model that learns schemas from unlabeled text without huma
 n supervision. It achieved state of the art performance on an information 
 extraction domain. Further\, the talk will briefly summarize the latest wo
 rk in the area\, and describe a new dataset to motivate future research in
  event schema learning.
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR