BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Learning to Adapt in Dialogue Systems: Data-driven Models for Pers
 onality Recognition and Generation - Speaker to be confirmed
DTSTART:20081121T120000Z
DTEND:20081121T130000Z
UID:TALK14800@talks.cam.ac.uk
CONTACT:Johanna Geiss
DESCRIPTION:Most dialogue systems do not take linguistic variation into ac
 count in both the understanding and generation phases\, i.e. the user's li
 nguistic style is typically ignored\, and the style conveyed by the system
  is chosen once for all interactions at development time. We believe that 
 modelling linguistic variation can greatly improve the interaction in dial
 ogue systems\, such as in intelligent tutoring systems\, video games\, or 
 information retrieval systems\, which all require specific linguistic styl
 es. Previous work has shown that linguistic style affects many aspects of 
 users' perceptions\, even when the dialogue is task-oriented. Moreover\, u
 sers attribute a consistent personality to machines\, even when exposed to
  a limited set of cues\, thus dialogue systems manifest personality whethe
 r designed into the system or not.\n\nOver the past few years\, psychologi
 sts have identified the main dimensions of individual differences in human
  behaviour: the Big Five personality traits. We hypothesise that the Big F
 ive provide a useful computational framework for modelling important aspec
 ts of linguistic variation. We explore the possibility of recognising the 
 user's personality using data-driven models trained on essays and conversa
 tional data. We then test whether it is possible to generate language vary
 ing consistently along each personality dimension in the information prese
 ntation domain. We present PERSONAGE: a language generator modelling findi
 ngs from psychological studies to project various personality traits. We u
 se PERSONAGE to compare various generation paradigms: (1) rule-based gener
 ation\, (2) overgenerate and select and (3) generation using parameter est
 imation models---a novel approach that learns to produce recognisable vari
 ation along meaningful stylistic dimensions without the computational cost
  incurred by overgeneration techniques. We also present the first human ev
 aluation of a data-driven generation method that projects multiple stylist
 ic dimensions simultaneously and on a continuous scale. 
LOCATION:SW01\, Computer Laboratory
END:VEVENT
END:VCALENDAR