Data is usually meta-data (data about data) - what a document is about, additional information like who the author is, etc. But, it can also be the raw information - like a business vocabulary.
Linkages are the relationships between the data. The information that ties the data together and lets you infer and extrapolate.
Infrastructure is the formalisms of the languages (RDF, RDF Schema, OWL, SPARQL, ...) and the services that are already provided (W3C's Linked Data, Protege, Pellet, ...). Data without backing services and formalisms means that you have to create everything yourself and there is no exponential building of knowledge that comes from sharing the data.
That's it. Let me know if you agree with my 3 words or have different ones.