Stories are a ubiquitous part of the human experience and an essential component of human intelligence. Stories serve countless functions working in tandem with our language facilities to enable communication, problem solving, imagination, and more. Stories are eminently cross disciplinary, fueling how people think about everything from chemical reactions and novels to social encounters and politics. If we’re to develop a comprehensive computational system mirroring human intelligence, the system must be able to both understand stories and go beyond that, using stories to empower how it thinks about the world.
Working to take computational story understanding to the next level, I have made a number of contributions centered on story generation. In particular, I have developed a novel story generation system that is capable of creating new and interesting stories by learning from a human written corpus of stories. In order to facilitate this development, I designed robust, learnable character models. The character models are capable of handling characters that span a number of genres including romantic comedies, fairy tales, warfare, and Shakespearean tragedies. I created methods for automatic learning of characters traits in both an unsupervised way using topic modelling and a semi-supervised way using alignment-based learning.
Through my research I’ve been able to test these algorithms using the Genesis system for story understanding. The results show how the methods I’ve developed are robust – able to handle internally conflicted characters, characterization spanning genres, and conflicting character goals. Critically, I demonstrate how story generation is important to story understanding through an example of using generation to reinterpret a known story. Finally, I create and discuss a tool to enable crowd sourced evaluation of both my own work and future experiments in computational story understanding.
Committee: Patrick Winston, Boris Katz, Nick Montfort