Inferring process models from temporal data with abduction and induction

This paper shows how automated abduction and induction can be used to infer logical process models from temporal observations of states and actions. The proposed method employs a non-monotonic learning system called eXtended Hybrid Abductive Inductive Learning (XHAIL) to learn domain axioms in a temporal logic programming formalism known as the Event Calculus (EC). The key benefit of this logical learning method is its ability to utilise background knowledge and to return human understandable hypotheses. The approach is illustrated on a simplified biological process modelling task.