BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Numerically Grounded Language Models - George Spithourakis (UCL)
DTSTART:20161021T110000Z
DTEND:20161021T120000Z
UID:TALK67250@talks.cam.ac.uk
CONTACT:Kris Cao
DESCRIPTION:Assisted text input and editing tools can save time and effort
  and improve text quality. For example\, word prediction presents the user
  with a list of choices for the next word\; word completion helps them com
 plete the word that they have started typing\; error detection identifies 
 erroneous spans of text\; error correction is recommending amendments to t
 he text. Traditional language models fail to adequately address such task 
 in domains that are rich in numerical mentions (e.g. clinical reports).\n\
 nIn this talk\, I discuss extensions to neural language models that are se
 nsitive to numerical values encountered in the text itself. Continuous num
 erical mentions provide exact measurements of the state of the world witho
 ut mapping to discrete verbal categories. Therefore\, the resulting models
  are grounded through numbers. The proposed framework can also be used for
  language modeling conditional on structured knowledge bases with inconsis
 tent schemas or missing attributes of variable types. Finally\, I will pre
 sent experimental results for text input and editing tasks where numerical
 ly grounded and conditional models yield state-of-the-art results.
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR