BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:NLIP reading group: Semi-Supervised Recursive Autoencoders for Pre
 dicting Sentiment Distributions - Diarmuid Ó Séaghdha (University of Cam
 bridge)
DTSTART:20111110T120000Z
DTEND:20111110T130000Z
UID:TALK34565@talks.cam.ac.uk
CONTACT:Jimme Jardine
DESCRIPTION:Diarmuid will be covering the following:\n\nRichard Socher\; J
 effrey Pennington\; Eric H. Huang\; Andrew Y. Ng\; Christopher D. Manning 
 Semi-Supervised Recursive Autoencoders for Predicting Sentiment Distributi
 ons \n\nhttp://aclweb.org/anthology-new/D/D11/D11-1014.pdf\n\nWe introduce
  a novel machine learning framework based on recursive autoencoders for\ns
 entence-level prediction of sentiment label\ndistributions. Our method lea
 rns vector space\nrepresentations for multi-word phrases. In\nsentiment pr
 ediction tasks these representations outperform other state-of-the-art app
 roaches on commonly used datasets\, such as\nmovie reviews\, without using
  any pre-defined\nsentiment lexica or polarity shifting rules. We\nalso e
 valuate the model’s ability to predict\nsentiment distributions on a new
  dataset based\non confessions from the experience project.\nThe dataset c
 onsists of personal user stories\nannotated with multiple labels which\, w
 hen\naggregated\, form a multinomial distribution\nthat captures emotional
  reactions. Our algorithm can more accurately predict distributions over s
 uch labels compared to several\ncompetitive baselines.
LOCATION:GS15\, Computer Laboratory
END:VEVENT
END:VCALENDAR