Dynamic distributional semantics
Sep 14-15, 2015, Mo/Tu
Organizers: Marco Baroni and Hinrich Schuetze
Original Topic Description: Dynamic semantics in formal semantics makes intuitive sense to me because "context-less" meaning representations are often needlessly complicated and often do not model form-meaning correspondence in an explanatory way. Examples include pronouns, scope and, in general, local ambiguity that disappears when taking larger context into account. This same argument can be made for distributional semantics. Is it really helpful to represent the meaning of a sentence as a vector? Or is a sentence better thought of as a function that takes one discourse state as argument and returns an updated discourse state?
More focused topic after initial discussions: Distributional semantics and reference