Is Information Density Uniform in Task-Oriented Dialogues?

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 8271–8283 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 8271 Is Information Density Uniform in Task-Oriented Dialogues? Mario Giulianelli, Arabella Sinclair, Raquel Fernández Institute for Logic, Language and Computation University of Amsterdam {m.giulianelli|a.j.sinclair|raquel.fernandez}@uva.nl Abstract The Uniform Information Density principle states that speakers plan their utterances to re- duce fluctuations in the density of the informa- tion transmitted. In this paper, we test whether, and within which contextual units this princi- ple holds in task-oriented dialogues. We show that there is evidence supporting the principle in written dialogues where participants play a cooperative reference game as well as in spoken dialogues involving instruction giving and following. Our study underlines the im- portance of identifying the relevant contextual components, showing that information content increases particularly within topically and ref- erentially related contextual units. 1 Introduction Due to production and perception errors, differ- ences between individuals, and other sources of uncertainty, language use for information transmis- sion can be thought to happen through a noisy chan- nel. Effective and efficient information exchange under such conditions can be modelled using the tools of Information Theory (Shannon, 1948). In- deed, information-theoretic models have success- fully accounted for surprisal in speech perception (Jelinek et al., 1975; Clayards et al., 2008), read- ing (Keller, 2004; Demberg and Keller, 2008; Levy et al., 2009), and sentence interpretation (Levy, 2008; Gibson et al., 2013), providing psycholin- guistic evidence that the information content of linguistic signals is related to comprehension pro- cessing effort. Speakers, too, are sensitive to the properties of the communication channel. They are thought to si- multaneously minimise their own production effort and the addressee’s processing effort (Clark and Wilkes-Gibbs, 1986; Clark and Schaefer, 1989). The most efficient way of dealing with both pres- sures, according to Information Theory, is to trans- mit information at a constant rate (Genzel and Char- niak, 2002), making linguistic choices that reduce fluctuations in the density of the information trans- mitted. Evidence for the principle of uniform in- formation density (UID; Jaeger and Levy, 2007; Jaeger, 2010) has been found at many levels of language production: speakers tend to reduce the duration of more predictable sounds (Aylett and Turk, 2004, 2006; Bell et al., 2003; Demberg et al., 2012); they tend to drop sentential material within more predictable scenarios (Jaeger and Levy, 2007; Jaeger, 2010; Frank and Jaeger, 2008); in spoken dialogue they are more likely to overlap at turn tran- sitions when information density is low (Dethlefs et al., 2016); and the rate at which they transmit in- formation in texts is uniform (Genzel and Charniak, 2002, 2003; Qian and Jaeger, 2011). Empirically, it is as yet unclear whether information density re- mains uniform throughout conversations (Vega and Ward, 2009; Doyle and Frank, 2015a,b; Xu and Reitter, 2018). That information density is not always uniform in dialogue may be due to the complex structure of conversational context (Clark and Brennan, 1991), which not only includes previous utterances …

Content truncated to 500 words