286 12 Building Bioinformatics Ontologies
- Graph-based languages.These include semantic networks and conceptual
graphs. Knowledge is represented using nodes and links between the
nodes. XML and the Semantic Web languages are the best-known exam-
ples of graph-based languages.
Perhaps because of the strong analogy between hypertext and semantic net-
works, most recent ontology languages have been graph-based.
Deciding what approach to use for building an ontology is not an easy one.
In this book, the emphasis is on the major web-based approaches as follows: - XML DTDis the most basic as well as the most widely supported. How-
ever, it has serious limitations as an ontology language. - XSDis quickly gaining acceptance, and conversion from XML DTD to
XSD has been automated. However, it shares most of the limitations of
XML DTDs. - XML Topic Mapsis a language for defining abstract subjects, calledtop-
ics, and the relationships between them. Topic maps directly support
higher-order relationships, which is not the case for the other languages
in this list. On the other hand, topic maps do not have the complex data
structures of XSD or the sophisticated semantics of RDF and OWL. Un-
fortunately, there are very few tools available for XML Topic Maps, so the
development of ontologies using this language will not be discussed in
this chapter. - RDFhas been gaining in popularity. There are fewer tools available for
RDF than there are for XML, but new tools are continually becoming
available. Unfortunately, there is no easy path for converting from XML
DTDs or schemas to RDF (or vice versa). RDF has some inference built
in, and RDF semantics is compatible with modern rule engines (either
forward- or backward-chaining). It is also well suited to high-performance
graph matching systems. - RDF specified with an XML DTDis an approach that is compatible with
XML DTD, XSD, RDF and the OWL languages. The Gene Ontology (GO)
has used this technique. In this approach the DTD is designed so that it
complies with RDF as well as with the OWL languages. Since an XML
DTD can easily be converted to XSD, this makes the documents compati-
ble with all major Web based ontology languages except for Topic Maps.
However, only the most rudimentary features of RDF and OWL can be
used by this approach.