Standardisation Action Plan for CLARIN

Knowledge Engineering p. 5 Recommendations p. 6

Suggested wording: CLARIN recommends to closely follow all the future developments in the W3C languages domain, and to make use of simple W3C standards like RDF(S) whenever possible – i.e., when complex knowledge representation and reasoning problems are not at stake. Various frameworks recommended in CLARIN should provide an export into RDF(S) format, whereas all export into OWL format should require a careful examination in terms of utility, simplicity of use and computational complexity.

Open issues p. 6

Comments: Interoperability is surely a strong argument in favour of the adoption of a W3C approach. However, because mainly of their ‘binary’ nature (for RDF(S) and OWL, properties are relations with only two arguments) W3C languages are subject to many practical limitations and have a limited ‘expressive’ power. This is may be not very important for RDF(S), planned from the beginning as a ‘low level’ (from a knowledge representation point of view) language – moreover, RDF(S) and its variants (RDFa) can probably find their utility in applications of the “browser of the future” type like the so-called “linked data”. With respect now to OWL (and OWL-2), its ‘practical’ utility has never been demonstrated (it is particularly insufficient from an expressiveness point of view, it does not support rules in a correct way, it is based on an ‘open world’ assumption that precludes its use in many industrial applications, it is verbose and, at the same time, too rigid because of its formalization in Description Logics terms, etc.).

TimeML p. 9

Comments on the main TimeML Section, valid also for the “Recommendations” and “Open Issues” Sections: On the other hand, TimeML is also a particularly (and unnecessarily) intricate and ‘heavy’ model that will probably never be used in its entirety. Simpler and more efficient schemata for temporal representation exist, e.g., for NL historical and narrative documents, like the ‘temporal’ component of NKRL (Narrative Knowledge Representation Language) - see at the address http://www.springer.com/computer/artificial/book/978-1-84800-077-3. The ‘temporal sub-language’ of NKRL concerns both the representation of the temporal phenomena, and their use for indexing, querying and inferencing applications.