Modelling Mutually Interactive Fictional Character Conversational Agents
Gert Verhulst recently announced that he would quit the TV show “Samson & Gert”. But fear not: we have already automated the entire cast of the show! In this talk, we present our novel probabilistic text generation model Babbly, our novel Twitterbot framework, and show how to make complex chatbot behaviors by implementing a multi-agent Twitterbot system based on six Samson & Gert characters, all constantly conversing with each other.
Thomas Winters is a PhD student at the KU Leuven DTAI research group, researching computational creativity and probabilistic models, with a particular interest in computational humor.
More specifically, the research is about how probabilistic logic learning can be applied to automatically learn efficient generative models within computational creativity.