Why discourse affects speakers’ choice of referring expressions

Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing, pages 1639–1649, Beijing, China, July 26-31, 2015. c⃝2015 Association for Computational Linguistics Why discourse affects speakers’ choice of referring expressions Naho Orita Graduate School of Information Sciences Tohoku University naho@ecei.tohoku.ac.jp Eliana Vornov Computer Science and Linguistics University of Maryland evornov@umd.edu Naomi H. Feldman Linguistics and UMIACS University of Maryland nhf@umd.edu Hal Daum´e III Computer Science and UMIACS University of Maryland hal@umiacs.umd.edu Abstract We propose a language production model that uses dynamic discourse information to account for speakers’ choices of refer- ring expressions. Our model extends pre- vious rational speech act models (Frank and Goodman, 2012) to more naturally dis- tributed linguistic data, instead of assuming a controlled experimental setting. Simula- tions show a close match between speakers’ utterances and model predictions, indicat- ing that speakers’ behavior can be modeled in a principled way by considering the prob- abilities of referents in the discourse and the information conveyed by each word. 1 Introduction Discourse information plays an important role in various aspects of linguistic processing, such as predictions about upcoming words (Nieuwland and Van Berkum, 2006) and scalar implicature process- ing (Breheny et al., 2006). The relationship be- tween discourse information and speakers’ choices of referring expression is one of the most studied problems. Speakers’ choices of referring expres- sions have long been thought to depend on the salience of entities in the discourse (Giv´on, 1983). For example, speakers normally do not choose a pronoun to refer to a new entity in the discourse, but are more likely to use pronouns for referents that have been referred to earlier in the discourse. A number of grammatical, semantic, and distribu- tional factors related to salience have been found to influence choices of referring expressions (Arnold, 2008). While the relationship between discourse salience and speakers’ choices of referring expres- sions is well known, there is not yet a formal ac- count of why this relationship exists. In recent years, a number of formal models have been proposed to capture inferences between speak- ers and listeners in the context of Gricean prag- matics (Grice, 1975; Frank and Goodman, 2012). These models take a game theoretic approach in which speakers optimize productions to convey in- formation for listeners, and listeners infer meaning based on speakers’ likely productions. These mod- els have been argued to account for human commu- nication (Jager, 2007; Frank and Goodman, 2012; Bergen et al., 2012a; Smith et al., 2013), and stud- ies report that they robustly predict various linguis- tic phenomena in experimental settings (Goodman and Stuhlm¨uller, 2013; Degen et al., 2013; Kao et al., 2014; Nordmeyer and Frank, 2014). However, these models have not yet been applied to language produced outside of the laboratory, nor have they incorporated measures of discourse salience that can be computed over corpora. In this paper, we propose a probabilistic model to explain speakers’ choices of referring expres- sions based on discourse salience. …

Content truncated to 500 words