[open-literature] Open Correspondence RDF
k.j.w.alexander at gmail.com
Fri Nov 5 03:37:04 GMT 2010
I have some suggested improvements for the RDF output of the Open
I have sketched out a modelling for one of the letters: http://bit.ly/axvkF2
(I am also CCing the "pedantic web" mailing list, whose members may
have some constructive feedback, and possibly pick up any mistakes
First of all, the URIs you use for the letters are the same URIs you
use for the web pages showing the letters. It would be really useful
to make new URIs for the letters so that, eg, you can clearly say that
Dickens wrote the letter, and you created the web page displaying the
In the sample RDF I sketched out, I used the URI
to identify the Letter.
Instead of the letter:correspondent property being "Miss%20Hogarth",
it's more useful if we use a URI to identify the person, that way we
can describe Miss Hogarth. I think it would be a nice idea to have
pages for each of the people written to and mentioned in the letters -
the pages could contain more data about the person, as well as links
to the letters that mention them. You could also do places and
I've also given the letter an rdf:type of letter:Letter, so that you
can tell what sort of thing it is. I notice that the namespace you
use for the "letter" schema is http://purl.org/letter, which doesn't
appear to be registered - so I think you should either register it if
you can (not sure you can have PURL names at the top level?), or maybe
dc:author doesn't exist, it's better to use dcterms:creator, and again
it's really useful to use a URI to identify the creator, not just a
name, because then you can have more data about the creator.
letter:textReferred - the letter I remodelled didn't refer to a text,
but again, it's good to have a URI to identify each text. I think
dcterms:subject would be an adequate property to use instead of
letter:textReferred, but inventing letter:textReferred is OK too.
I also think it's useful to have the body of the letter in the RDF as
well, not just the metadata (I think someone made this point before).
I have used sioc:content, but inventing something like letter:text
would also be an option.
I've picked out the opener, closer, addressline and textual date, with
properties I've pretended are in the letters schema. I guess they
could be useful to index on etc, and would not be too hard to parse
out of the letter?
I used dcterms:subject to link to some of the things mentioned in the
letter, and described some of them (eg, the people and their
relationships to each other) elsewhere in the RDF.
Keith Alexander (RDF pedant)
More information about the open-literature