Marc Mason and Carolyn Penstein Rose
In this paper we discuss an application of genetic programming to the problem of learning constraints to be used in plan operators for language interpretation. We use as an example the problem of identifying the discourse purpose of Okay. In particular, we compare a pure genetic programming approach with a hybrid genetic programming/plan-based approach. We demonstrate that genetic programming allows for a natural combination of machine learning with more traditional plan-based discourse processing approaches. The type of discourse that we focus on in this paper is humanhuman spontaneous negotiation dialogues.