Entity annotations

Our corpus contains annotated typed entity references. We annotate text strings as entity references when they refer to clearly distinguishable individual instances in the real or fictional world. This makes entity references highly related to mentions from coreference resolution, although there are some conceptual differences. We further annotated the type of the referred entity as part of the entity reference. In determining the type of the referred entity, we handle fictional worlds in the same way as the real world. In the absence of other indications, we assume that fictional worlds follow the same rules as the real world.

The types we distinguish are

  • Person: Reference to humans or humanly acting animals or things
  • Organisation: Reference to an organisation
  • Location: Reference to a place or location
  • Work: Reference to a cultural artefact (publication, law)
  • Event: Reference to an event
  • Concept: Reference to domain-specific theoretical concept

Please find more detailed information as well as examples and discussion of special cases in the annotation guidelines.

Named entities are a subset of the potentially referring expressions.