Linked data in packaged content (ePub)

I'm interested in putting linked data into eBooks published in the
(open standard) ePub format (http://www.openebook.org/ ). The format
is essentially a relocatable zip file of XHTML, associated media files
and a few metadata files.

The target platforms of this content impose some restrictions on what
is practical: e-ink devices (which are the only current eBook readers
with the battery life to last an entire novel) typically don't have an
internet connection (thus no resolving of links) and have very little
in the way of processing power (thus no full reasoning).

We already have some data-interlinking between our collection
(http://www.nzetc.org/ ) and librarything
(http://www.librarything.com/ ) at the FRBR work level
(http://vocab.org/frbr/core.html#Work ) and also some links to
wikipedia / dbpedia for named entities (principally authors and
places).  We believe we have quite good authority control over author
names, even those who published under multiple names (see, for example
http://www.nzetc.org/tm/scholarly/name-208662.html or
http://www.nzetc.org/tm/scholarly/name-208310.html ). We have ~1300
ePubs, the largest of which exceed the size limits of most ePub tools.

Does anyone know of any other attempts to put linked data into
packages like this?

There are two main issues I can see: (a) how to self-identify the
package (naive hashing doesn't work, as some eBook readers open the
package and add custom metadata) and (b) how to package the linked
data to get maximal use when a paucity of CPU precludes a full
reasoner.

The traditional identifier used in this field, the ISBN, is
essentially a print-run identifier, and not of a whole lot of obvious
use to us since: (a) most of our books' original publishing predates
ISBNs and (b) our digital republishing of them doesn't qualify for an
ISBN according to our local ISBN issuer (the National Library of New
Zealand).

cheers
stuart

Received on Wednesday, 28 April 2010 08:02:20 UTC