The Description and Processing of Multiword Expressions in OntoSem

The paper discusses the lexical description and runtime treatment of multiword expressions in the Ontological Semantic (On-toSem) text processing environment. We show how our formalism permits the encoding of syntactic structures and semantic entities of any degree of complexity, and how those descriptions support the automatic generation of high quality, formal text-meaning representations from unrestricted input text. We argue for the need for high quality, largely manually crafted knowledge if NLP applications are to become really sophisticated. We shift the focus from the automatic detection of MWEs (the most widespread interest of late) to the encoding of their meaning, since detection without meaning can support only shallow NLP applications.