As reported in the Semantic Web Activity Blog, a new Interest Group Note has just been published, documenting the extraction algorithm from HTML5+Microdata to RDF. The algorithm replaces the one that was part of an earlier version of the Microdata+HTML5 draft.
In the past few days I worked on an implementation of this extraction algorithm in Python; I have now created a public service at W3C to serve as an RDF distiller for microdata. It is still an alpha release, meaning that I am sure it has bugs; but it may be useful nevertheless. Comments are, of course, welcome!