Empirical verification of Meaning-Game-based generalization of centering theory with large Japanese corpus

Centering theory (Grosz et al., 1995) tries to explain relations among attention, anaphora, and cohesion. It has two theoretical limitations. The first is the lack of a principle behind these discourse phenomena. The second is that the salience of discourse entities has not been quantitatively defined, although it plays a critical role in this theory. Hasida et al. (1995, 1996) propose the meaning game as a more principled model of intentional communication based on game theory, and claim that it can derive centering theory. This claim, however, has not yet been verified on the basis of substantial linguistic data. In this paper, we formulate salience as a measurable quantity in terms of a reference probability. We also formulate preferences subsuming centering theory under this quantitative formulation of salience. The preferences are derived from the meaning game and entail more general predictions than those of conventional centering theory. These formulations overcome the above limitations of centering theory. By following them, we empirically verify our generalization with a large Japanese corpus. The experimental results show that there is positive correlation between the salience (reference probability) of an entity and the simplicity (utility) of a noun phrase which refers to the entity. They also indicate correspondence between the values of expected utility and the ranking of the transition states. These results indicate that our generalization is appropriate.