REGRESSION A Dissertation

ACKNOWLEDGMENTS Charles Sutton told me that graduate school is supposed to be a life changing experience. He was correct. More people have helped me through this process than I can list here, but several deserve particular mention. I am here because of all of my teachers. Lisa Meeden at Swarthmore introduced me toMiller taught me to reason about uncertainty. Other graduate students and postdocs at UMass and elsewhere have been a vital part of my education, but the members of the IESL group have been particularly helpful. Shorey created a wonderful environment to learn, practice, and teach. Hanna Wallach was a constant source of inspiration. Our collaboration is my model of an effective research team. My work for the Perseus Project at Tufts University prepared me for life as a graduate student. Nobody reaches orbit without a big rocket behind them, and Gregory Crane has been a Saturn V. I chose the members of my committee for their excellence, and I am pleased that they have demanded the same from me. I have known David Smith the longest, since I undertook the difficult task of filling the void he left at the Perseus Project. David Jensen's wisdom has been vital. Michael Lavine arrived, as if summoned, at precisely the time when he made the biggest impact on how I understand statistics and how I explain it to others. David Blei has been an inspiration even before I began graduate school. I depend on his skill and creativity every day. v Andrew McCallum took an enormous risk on an untested student with an odd background and not much math, and has backed me 100% throughout this process. I hope I have repaid his trust. Owen is almost as old as my graduate career. Nathaniel was born around when I began working on this thesis. Watching them grow and learn has given me perspective on my own growth as a computer scientist. Graduate school is one of many things that I could not have done without Jennifer's constant support and love. My mother, Nancy, taught me to value creativity, and it has become my most important asset. My father, Gerry, was the first person who told me what a PhD was, and the first to suggest that I might like to get one. vi ABSTRACT TOPIC REGRESSION in text analysis has often involved predicting these non-text values based on text data for …