Generating Dialogue from
Personality-Ranked Utterances
Carsten Brockmann
Previous research has shown that
users prefer human-computer interfaces that match their own personality
traits (Nass and Lee 2001). Based on studies on personality and
language use (Pennebaker and King 1999, Gill and Oberlander
2002--2004), we identified a set of textual features that
indicate an utterance's underlying personality. We now want to
apply these features for the generation of text with
recognizable personality.
In this talk, I will present a framework for generating dialogue
in which two computer characters discuss a movie. The
characters are parameterized for personality, and their utterances are
chosen from a pre-defined corpus. Content and surface
form of potential utterances are ranked for personality,
making use of the FLIGHTS user modeling component (Foster and White
2003).
This PhD work is part of the Edinburgh-Stanford Link project
Critical Agent Dialogue (CrAg).