Unsupervised Induction of Contingent Event Pairs from Film Scenes

Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 369–379, Seattle, Washington, USA, 18-21 October 2013. c⃝2013 Association for Computational Linguistics Unsupervised Induction of Contingent Event Pairs from Film Scenes Zhichao Hu, Elahe Rahimtoroghi, Larissa Munishkina, Reid Swanson and Marilyn A. Walker Natural Language and Dialogue Systems Lab Department of Computer Science, University of California, Santa Cruz Santa Cruz, CA, 95064 {zhu, elahe, mlarissa, reid, maw}@soe.ucsc.edu Abstract Human engagement in narrative is partially driven by reasoning about discourse relations between narrative events, and the expectations about what is likely to happen next that results from such reasoning. Researchers in NLP have tackled modeling such expectations from a range of perspectives, including treating it as the inference of the CONTINGENT discourse relation, or as a type of common-sense causal reasoning. Our approach is to model likeli- hood between events by drawing on several of these lines of previous work. We implement and evaluate different unsupervised methods for learning event pairs that are likely to be CONTINGENT on one another. We refine event pairs that we learn from a corpus of film scene descriptions utilizing web search counts, and evaluate our results by collecting human judg- ments of contingency. Our results indicate that the use of web search counts increases the av- erage accuracy of our best method to 85.64% over a baseline of 50%, as compared to an av- erage accuracy of 75.15% without web search. 1 Introduction Human engagement in narrative is partially driven by reasoning about discourse relations between nar- rative events, and the expectations about what is likely to happen next that results from such reason- ing (Gerrig, 1993; Graesser et al., 1994; Lehnert, 1981; Goyal et al., 2010). Thus discourse relations are one of the primary means to structure narrative in genres as diverse as weblogs, search queries, sto- ries, film scripts and news articles (Chambers and Jurafsky, 2009; Manshadi et al., 2008; Gordon and Swanson, 2009; Gordon et al., 2011; Beamer and Girju, 2009; Riaz and Girju, 2010; Do et al., 2011). DOUGLAS QUAIL and his wife KRISTEN, are asleep in bed. Gradually the room lights brighten. the clock chimes and begins speaking in a soft, feminine voice. They don’t budge. Shortly, the clock chimes again. Quail’s wife stirs. Maddeningly, the clock chimes a third time. CLOCK (continuing)Tick, tock –. Quail reaches out and shuts the clock off. Then he sits up in bed. He swings his legs out from under the covers and sits on the edge of the bed. He puts on his glasses and sits, lost in thought. He is a good-looking but conventional man in his early thirties. He seems rather in awe of his wife, who is attractive and rather off-hand towards him. Kirsten pulls on her robe, lights a cigarette, sits fishing for her slippers. Figure 1: Opening Scene from Total Recall Recent work in NLP has tackled the inference of relations between events from a broad range of per- spectives: (1) as inference of a …

Content truncated to 500 words