Guiding Interactive Film With Emotion-Profiling Chatbots

Celina L. Ma, Haohong Wang, Mea Wang

Abstract: In interactive films, users have chances to influence the storyline at fixed points in time. To grant more user autonomy, this work offers a new model for interaction, driven by chatbots representing story characters. Whenever they desire, users may interact with the film characters in open-ended conversation. Through generative language models, the chatbots can respond according to the story context for dynamic entertainment. This unique interface is augmented by emotion recognition to profile the users’ character preferences, and guide them to suitable storylines. A user study on a prototype film confirmed the subjective value of this system, and its potential to drive story progression with minimal authorial scripting. An objective analysis of path selection algorithms further showed the benefits of assessing long-term storyline quality.