Non-deterministic attribute selection in reference production

Abstract

In producing identifying descriptions, speakers often overspecify and manifest preferences for certain at-tributes. However, current computational models which incorporate this observation tend not to make precise predictions about when and how much speakers do this. The present paper proposes and evaluates two alter-native models, based on the results of a new experi-ment. Unlike current models, the new ones are non-deterministic and seek to make precise quantitative pre-dictions about the extent to which speakers overspecify.