Right now, we are handling Darwin Core record-level terms by assigning a subset
of them as properties of Occurrence, and simply ignoring the others (i.e.,
those that don't make sense as properties of Occurrence).
This is a straightforward solution that seems to work well enough for most
real-life examples of DwC data, but it has several shortcomings.
1) Some record-level terms are dropped.
2) There could be a lot of redundancy in the triples because many (or all)
records in a DwC dataset often share the same values for the record-level terms.
A possible solution to both 1) and 2) would be to define a new "metadata" class
that has the record-level terms as its properties. Then, instances of this
metadata class could be assigned as a property of Occurrence instances. This
would allow us to handle all record-level terms and also avoid (or at least
minimize) the redundancy problem.
The downside is that this requires constructs (e.g., the metadata class) that
are currently not a part of DwC. We feel that for now, it makes the most sense
to stick with the initial solution described above, but this should be
revisited in the future as a possible enhancement.
Original issue reported on code.google.com by stucky.b...@gmail.com on 5 Apr 2013 at 5:00
Original issue reported on code.google.com by
stucky.b...@gmail.com
on 5 Apr 2013 at 5:00