Multiple Narrative Disentanglement: Unraveling Infinite Jest

2012 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1–10, Montr´eal, Canada, June 3-8, 2012. c⃝2012 Association for Computational Linguistics Multiple Narrative Disentanglement: Unraveling Infinite Jest Byron C. Wallace Tufts University and Tufts Medical Center Boston, MA byron.wallace@gmail.com Abstract Many works (of both fiction and non-fiction) span multiple, intersecting narratives, each of which constitutes a story in its own right. In this work I introduce the task of multiple nar- rative disentanglement (MND), in which the aim is to tease these narratives apart by assign- ing passages from a text to the sub-narratives to which they belong. The motivating exam- ple I use is David Foster Wallace’s fictional text Infinite Jest. I selected this book because it contains multiple, interweaving narratives within its sprawling 1,000-plus pages. I pro- pose and evaluate a novel unsupervised ap- proach to MND that is motivated by the theory of narratology. This method achieves strong empirical results, successfully disentangling the threads in Infinite Jest and significantly outperforming baseline strategies in doing so. 1 Introduction Both fictional and non-fictional texts often com- prise multiple, intersecting and inter-related narra- tive arcs. This work considers the task of identifying the (sub-)narratives latent within a narrative text and the set of passages that comprise them. As a mo- tivating example, I consider David Foster Wallace’s opus Infinite Jest (Wallace, 1996),1 which contains several disparate sub-narratives interleaved through- out its voluminous (meta-)story. By sub-narrative I mean, loosely, that these threads constitute their own independent stories, coherent on their own (i.e., 1No relation. without the broader context of the overarching narra- tive). I refer to the task of identifying these indepen- dent threads and untangling them from one another as multiple narrative disentanglement (MND). The task is of theoretical interest because disen- tanglement is a necessary pre-requisite to making sense of narrative texts, an interesting direction in NLP that has received an increasing amount of atten- tion (Elson et al., 2010; Elson and McKeown, 2010; Celikyilmaz et al., 2010; Chambers and Jurafsky, 2008; Chambers and Jurafsky, 2009). Recogniz- ing the (main) narrative threads comprising a work provides a context for interpreting the text. Disen- tanglement may thus be viewed as the first step in a literary processing ‘pipeline’. Identifying threads and assigning them to passages may help in auto- matic plot summarization, social network construc- tion and other literary analysis tasks. Computational approaches to literature look to make narrative sense of unstructured text, i.e., construct models that relate characters and events chronologically: disentangle- ment is at the heart of this re-construction. But MND is also potentially of more pragmatic import: disentanglement may be useful for identify- ing and extracting disparate threads in, e.g., a news- magazine article that covers multiple (related) sto- ries.2 Consider an article covering a political race. It would likely contain multiple sub-narratives (the story of one candidate’s rise and fall, a scandal in a political party, etc.) that may be of interest indepen- dently of the particular race at …

Content truncated to 500 words