Data-Driven Creativity Enhancement Through Word Association

Abstract

Writing creative stories from a blank page is a challenging task, particularly in the modern game industry where dozens of writers can contribute to the stories and settings of a constantly evolving artificial world. We introduce a system which recommends interesting, evocative, and thematically coherent words to help creators write thematically connected stories. We combine principles from human creativity enhancement and computational creativity to build a creative assistant based on word association research. We show that careful corpus selection, filtering based on emotional sentiment, and promoting remote associations through paragraph scale segmentation can produce recommendations that promote creative goals better than alternative word association algorithms according to our creative word indicators.iii I would like to begin by thanking my thesis supervisor, Dr. David Mould. His attention to detail, his patience, and his enthusiasm for this work was essential during this research. Regardless of the circumstances, his positive attitude and passion in our meetings would always leave me reinvigorated and ready to get back into the work. I could not have completed this work without his outstanding kindness, advice, patience, and support. His guidance has shaped my way of looking at computer science since my first year as an undergraduate student and has kept me driven to pursue this research.I would like to thank the thesis committee for reviewing my work and providing me with valuable suggestions. Their input has improved this thesis and has helped me improve as a researcher. I would also like to acknowledge Carleton University, the School of Computer Science, and GIGL for their financial support and thank them for all of the amazing opportunities they have afforded me over these years. I would like to thank all of my friends, colleagues, and visiting scholars in the GIGL lab for their advice on this thesis, on research, and in academics. iv iv Table of Contents vList of Tables viii List of Figures x4.12 Multi-stimuli recommendations when taking the intersection of TF-IDF weightings using the themes in Table 4.9. Ten recommendations were attempted, but some were unable to find enough recommendations when only looking at the top