A maiden's dream "The most sophisticated (and the most bleeding edge) is to use the Web Ontology Language, which allows you to define an ontology. Literally, "ontology" is the science or study of being, particularly in a metaphysical sense. It is not therefore, numerable and should not be preceded by the indefinite article...Now, the Web Ontology Language is a W3C (World Wide Web Consortium) standard and, in practice, it is an extension to UML. This gives the clue to how you create an ontology: you model it.
This also gives an indication of Unicorn's approach and why it tries not to talk about semantics too much. Unicorn aims to provide your computers with the ability to understand that customers and clients are the same thing but does so without forcing you to adopt a formal semantic approach. Thus, for example, you can do the relevant modelling using E-R (entity relationship) diagrams if you wish."
The second article covers Network Inference: "Network Inference is much more committed to new developments in the area of semantic information architectures, whereas Unicorn takes a more evolutionary approach. In particular, Network Inference is strongly committed to OWL, the Web Ontology Language (whose letters don't match because that is how Owl spelled his name - WOL - in Winnie the Pooh)...The big argument in favour of this approach is that OWL is a standard. So are XQuery, SOAP, UDDI, WSDL and the other standards that Network Inference supports. In principle, this is a significant step forward in the integration space (whether EAI or EII) because, while existing solutions make use of standards, they are not, with the arguable exception of pure XML solutions, standards-based. That is, there is no standard way of expressing a virtual schema or database to underpin these integrations."