In the past, semantic web technologies do not really fit on web pages due to the scarce methods of integration between ontologies and html pages. HTML microdata mechanism allows machine-readable data to be embedded in HTML documents in an easy-to-write manner, with an unambiguous parsing model. It is compatible with numerous other data formats including RDF and JSON. It can attach concept of a given taxonomy/ontology to words, paragraphs or complete HTML pages in a machine-readable format allowing for further inference.
In case you can to read the current draft, just see HTML 5 microdata W3C specification.