David Schlangen (Potsdam)
Concepts, Composition, and Conversational Coordination: Semantic Competence for Situated Interaction
In this lecture series, I will present some results from a research programme that my group has been following for some years now. The main question is what the semantic competence is that is required for talking about objects and situations that are directly perceivable, as well as those that have been observed, but aren't anymore at the time of talking. I will deal with this question from the perspective of modelling this competence in an artificial agent.
The starting idea is that this involves a set of core competences that includes at least: the competence to apply natural language expressions to the perceived situation; the competence to name objects, relations, and situations that are currently perceived, for others; the competence to relate what was said to other expressions. These core competences can be distinguished with respect to learning, representation, and application, but they come together in the phenomenon of reference, both in direct, singular reference and in discourse reference.
The focus of the lectures will be on computational experiments that we did to evaluate our implementation of this "reference/inference" model, but I will also discuss potential theoretical implications of the model, and will try to situate the proposal in the literatures on human word meaning acquisition and on situated semantics, and touch upon issues in the philosophy of language related to singular reference.