Spatial Relation Extraction using Relational Learning

The automatic extraction of spatial information is a challenging and novel task with many applications. We motivate our definition of this task and formulate it as an information extraction step prior to mapping to spatial semantics. Each sentence gives rise to several spatial relations between words representing landmarks, trajectors and spatial indicators. Learning to extract such spatial relations can be formulated as a typical relational classification problem, for which we employ the recently introduced kLog framework. We discuss modeling and representation, and show experimental results.