A methodological note on the definition of semantic annotation languages (short paper)

In the last few years, the international organization for standards ISO has started up various projects concerned with the definition of interoperable concepts for syntactic, morphosyntactic, and semantic annotation, with the ultimate aim to support the development of interoperable language resources. The Linguistic Annotation Framework (LAF, Ide & Romary, 2004) thereby serves as a meta-framework. LAF distinguishes between the concepts of annotation and representation: ‘annotation’ refers to the process of adding information to segments of language data, or to that information itself, independent of the format in which this information is represented. The term ‘representation’ refers to the format in which an annotation is rendered, for instance in XML. According to LAF, annotations are the proper level of standardization. This distinction is reflected in the specification of ISO-TimeML, a proposed ISO standard for temporal annotation (ISO, 2008) which consists of an abstract syntax, a concrete syntax, and a semantics. The abstract syntax specifies the elements making up the information in annotations, and how these elements may be combined to form complex annotation structures; these combinations are defined as set-theoretical structures. The concrete syntax is a variant of the TimeML markup language (Pustejovsky et al., 2003). Any other representation that is a rendering of the abstract syntax can be converted into this representation. The ISO-TimeML semantics is associated with its abstract syntax, which explains why all concrete representations of ISO-TimeML annotations are semantically equivalent. In this note we argue that the distinction of an abstract and a concrete syntax level is desirable not only from a standardization point of view, but also for designing annotation languages with a representation that is conceptually transparent for annotators and that allows a simple, systematic