BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Employing Psycholinguistics to Understand Decoding in Probabilisti
 c Language Generators - Clara Meister\, ETH Zurich
DTSTART:20240209T120000Z
DTEND:20240209T130000Z
UID:TALK211384@talks.cam.ac.uk
CONTACT:Richard Diehl Martinez
DESCRIPTION:Standard probabilistic language generators often fall short wh
 en it comes to producing coherent and fluent text despite the fact that th
 e underlying models perform well under standard metrics\, e.g.\, perplexit
 y.  This discrepancy has puzzled the language generation community for the
  last few years. In this talk\, we’ll take a different approach to looki
 ng at generation from probabilistic models\, pulling on concepts from psyc
 holinguistics and information theory in the attempt to provide insights in
 to some observed behaviors\, e.g.\, why high-probability texts can be dull
  or repetitive. Humans use language as a means of communicating informatio
 n\, aiming to do so in a simultaneously efficient and error-minimizing man
 ner\; in fact\, psycholinguistics research suggests humans choose each wor
 d in a string with this subconscious goal in mind. We propose that decodin
 g from probabilistic models of language should attempt to mimic these beha
 viors. To motivate this notion\, we’ll look at common characteristics of
  several successful decoding strategies\, showing how their design allows 
 them to implicitly adhere to attributes of efficient and robust communicat
 ion. We will then propose a new decoding strategy\, with the explicit aim 
 of encoding these human-like properties of natural language usage into gen
 erations.\n\nJoin Zoom Meeting\n\nhttps://cam-ac-uk.zoom.us/j/86071371348?
 pwd=OVlqdDhZNHlGbzV5RUZrSzM1cUlhUT09\n\n \n\nMeeting ID: 860 7137 1348\n\n
 Passcode: 387918\n\n
LOCATION:https://cam-ac-uk.zoom.us/j/86071371348?pwd=OVlqdDhZNHlGbzV5RUZrS
 zM1cUlhUT09
END:VEVENT
END:VCALENDAR