Skip to main content
A Script for Learning Scripts: Automatic Acquisition of Procedural Knowledge in Ontologically Grounded Agents
Abstract
Scripts describe events with component subevents. Ontologically grounded AI systems use scripts to represent the procedural knowledge necessary to execute complex tasks. In this paper, we present recent advances in dialogue-based script learning by agents configured in an ontologically grounded cognitive architecture. Specifically, we describe the procedure that details the sequence of events agents undertake when learning scripts in some subject domain or application. We demonstrate this script-learning script in a simulation system by tracing how an agent learns to replace a fuse through dialog with a human partner.