Mapping Personality to the Appearance of Virtual Characters Using Interactive Genetic Algorithms.
Résumé
We present an architecture for the generation of believable virtual characters from their personality traits. Using a crowd-based approach, we infer how people’s beliefs relate personality profile to physical appearance. The method we propose uses interactive genetic algorithms. Combined with natural language processing and sentiment analysis, it has the potential to drive the generation of new virtual characters from written description. In video games and online virtual worlds, it might replace traditional slider-based character personalization approaches by more creative, story-driven ones.