The inferences we draw from natural language meanings are beset with myriad sources of uncertainty. These sources include, among other things, vagueness, background knowledge, and the values of referring expressions like pronouns; on top of such sources, there is also uncertainty about the meanings of the expressions themselves (e.g., polysemy). Probabilistic Dynamic Semantics (PDS) provides a general framework for understanding uncertainty and context sensitivity in discourse by seamlessly integrating probabilistic effects with dynamic semantic analyses of lexical and discourse phenomena in the Montagovian tradition.

But it also has a second prong: it allows formal semanticists to construct probabilistic models of linguistic inference datasets by inferring the structure of uncertainty directly from data. Moreover, because it ties semantic theory to probabilistic models of linguistic data, it allows semanticists to compare alternative candidate theories of some phenomenon by assessing their relative fits to a given dataset.

Logo

Project Members

Papers