We propose PLANL, an approach for PLAnning with Natural Language, to accelerate the development of automated planning systems, enable plan sharing across multiple planners, and facilitate natural language interaction. PLANL uses generative sublanguage ontologies (GSOs) to robustly and accurately translate planning knowledge descriptions into representations such as STRIPS or hierarchical task networks. GSO’s accomplish this through a novel ability for efficiently representing and resolving polysemy. Unlike alternative approaches, PLANL does not have a proprietary plan representation. Instead, it exploits existing plan representations and selects a linguistically motivated conceptual vocabulary for them.