BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Factorising AMR generation through syntax - Kris Cao\, DeepMind
DTSTART:20190510T110000Z
DTEND:20190510T120000Z
UID:TALK121600@talks.cam.ac.uk
CONTACT:Andrew Caines
DESCRIPTION:Generating from Abstract Meaning Representation (AMR) is an un
 derspecified problem\, as many syntactic decisions are not constrained by 
 the semantic graph. To explicitly account for this underspecification\, we
  break down generating from AMR into two steps: first generate a syntactic
  structure\, and then generate the surface form. We show that decomposing 
 the generation process this way leads to state-of-the-art single model per
 formance generating from AMR without additional unlabelled data. We also d
 emonstrate that we can generate meaning-preserving syntactic paraphrases o
 f the same AMR graph\, as judged by humans.
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR