Modeling abduction over acyclic first-order logic horn theories in answer set programming: Preliminary experiments

Abstract

We describe encodings in Answer Set Programming for abductive reasoning in First Order Logic in acyclic Horn theories in the presence of value invention and in the absence of Unique Names Assumption. We perform experiments using the Accel benchmark for abductive plan recognition in natural language processing. Results show, that our encodings cannot compete with state-of-the-art realizations of First Order Logic abduction, mainly due to large groundings. We analyze reasons for this bad performance and outline potential future improvements.