Typical use of quantifiers: A probabilistic speaker model
Michael Franke · eScholarship (California Digital Library) · 2014
Many natural language quantifiers are classically associated with a stringent binary semantics, and similarly categorical pragmatic enrichments.But much experimental data shows that intuitions about appropriateness of use seem to be more fuzzy and more subtle, yet highly regular nonetheless.To account for these gradient typicality judgements, I sketch a new probabilistic model of Gricean speakers that incorporates a gradient notion of utterance alternatives.Focusing on scalar quantifier some, the model is a proof of concept, against expressed views to the contrary, that typicality and scalar inferences can be treated on a par.