Human Language Understanding: General Picture and Approaches, Sean Moens, Univ. Leuven, Belgium

Starting from the first attempts of automated language understanding we explain the most important state-of-the art approaches. We illustrate the latter with our research of translating narratives written in natural language to events happening in a 3D virtual world (FP7 FET project MUSE) and pinpoint pertinent bottlenecks. In a second part, we go deeper into important research questions of what representations are most suited, how to ground the representations in the real perceptual world, what properties machine learning models should have, and how to measure the success of language understanding. Finally, we discuss progress made in the MUSTER project (EU CHIST-ERA) towards these goals.