BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Decoding is deciding under uncertainty — the case of NMT - Bryan
  Eikema (University of Amsterdam)
DTSTART:20221111T120000Z
DTEND:20221111T130000Z
UID:TALK192329@talks.cam.ac.uk
CONTACT:Michael Schlichtkrull
DESCRIPTION:Abstract: \n\nIn neural machine translation (NMT)\, we search 
 for the mode of the model distribution to form predictions. We do so mostl
 y following the intuition that the most probable outcome ought to be an im
 portant summary of the distribution. Despite our intuition\, there’s ple
 nty of evidence against the adequacy of the most probable translations in 
 NMT. In this talk\, I make a case to move away from mode-seeking search as
  a tool for decision making as well as for model criticism. I will highlig
 ht reasons concerning MT as a task\, NMT as a probabilistic model\, and ML
 E as training algorithm. Finally\, I’ll turn to statistical decision the
 ory and motivate a different rule for making decisions\, one which is fami
 liar to statistical MT folks like those of my generation and earlier\, as 
 well as a modern approximation of it. I’ll close the talk with a discuss
 ion of merits and limitations of this decision rule\, and comments on oppo
 rtunities moving forward with or without mode-seeking search. \n\nBio: \n\
 nBryan Eikema is a PhD student at the Institute for Logic\, Language\, and
  Computation at the University of Amsterdam. His interests lie at the inte
 rsection of natural language processing and probabilistic modelling. In pa
 rticular he works on inducing latent structure in parallel data and improv
 ing neural machine translation through better probabilistic modelling. His
  research is part of the European GoURMET project. His thesis adviser is d
 r. Wilker Ferreira Aziz.\n\nPlease note: The speaker for this talk has bee
 n changed due to covid\; the content should be the same.\n\nTopic: NLIP Se
 minar\nTime: Nov 11\, 2022 12:00 PM London\n\nJoin Zoom Meeting\nhttps://c
 l-cam-ac-uk.zoom.us/j/91073515866?pwd=UnJmTER6dmZLeWpPOUo0VUJBOGxYQT09\n\n
 Meeting ID: 910 7351 5866\nPasscode: 646960\n\n
LOCATION:Virtual (Zoom)
END:VEVENT
END:VCALENDAR