Using inheritance and coreness sets to improve a verb lexicon harvested from FrameNet

We investigate two aspects of the annotation scheme underlying the FrameNet semantically annotated corpus — the inheritance relation on semantic types with its corresponding links between semantic roles of increasing granularity, and the specification of coreness sets of related semantic roles — against the background of our ongoing effort to harvest a lexicon of verb entries for deep parsing. We conclude that these aspects of the FrameNet annotation scheme do prove useful for reducing the complexity and ambiguity of verb entries, allowing for semantic roles of lower granularity for purposes of deep parsing, but need to be applied more systematically to make the lexicon usable in a practical parsing system.