Modeling prominence constraints for German pronouns as weighted retrieval cues

Anonymous

Abstract: We propose a cue-based retrieval model of German personal and demonstrative pronouns. The model extends the existing cue-based retrieval model of pronoun resolution by adding prominence constraints as weighted retrieval cues. We used data from an antecedent selection task reported in Schumacher et al. (2016). The experiment varied word order (canonical vs. non-canonical) and verb types (active accusative vs. dative experiencer) to test the effect of varying referential prominence on antecedent preferences for personal and demonstrative pronouns. Effectively, the model demonstrates that pronoun-antecedent dependency can be modeled as a cue-based retrieval process and the contrastive antecedent preferences of personal and demonstrative pronouns can be captured using weighted retrieval cues.