Since someone was asking about RDFa in the discussion:
I added a bit more explanation of how my python code works for those
that might be interested:
Even simply adding a triple or two per reference can be useful;
allowing one, say, to extract:
… and so to create a graph of citation references in a document.