Abstract

Rich annotation of linguistic resources is impeded because the de-facto annotation standard XML does not offer sufficient expressiveness. There is thus a trend towards defining super-models for linguistic annotation aiming to capture all possible requirements and allowing different representations for purposes of annotating, sharing, processing and querying. This paper argues that a mapping between these representations requires splitting rich annotations into layers each belonging to a structural class (list, map, tree, etc.). It is shown how these layers facilitate the crelat of an annotation tool suite using state-of-the-art components.