Towards a Reference Annotation Framework

This paper discusses the main characteristics of a possible unified framework for specifying annotation schemes dedicated to the task of reference identification and linking on linguistic corpora. Built upon the foundation principles of the Linguistic Annotation Framework, the model (RAF, Reference Annotation Framework) is based on the combination of a simple meta-model (expressing markables and links between them) and a selection of data categories representing the information actually attached to each component of the meta-model. Based on the observation of existing practices we show how this model can be used in a variety of practical and theoretical configurations.