> Please let us know the good the bad and the ugly as feedback from the
> field is very valuable to use (you know, when you stare at something for
> too long you get a little blind so we are not as good at improving it at
> this point).

So far I've only done one experiment with a largish data set (all RDF
generated from the arXiv's OAI export, scraped from arXiv article
index.html pages, or from citebase's OAI export). A great strength of
faceted browsing for those generating RDF metadata from other sources
is that it's often easy and painless to find unexpected problems with
your metadata. For example, I discovered lots of problems with my
parsing of comma separated fields into separate RDF statements, and
also found some strange bugs with reduplicated data coming from the
aXiv.

My biggest piece of feedback so far would really be that you need some
better documentation! Nothing is too hard once you've poked around a
bit, but since it is in principle so easy to get up and going, it
would be nice to have sufficient instructions to do so. (Things like:
it's using sesame for the rdf store, which files contain the rdf store
configuration, that it's useing lucene, where lucene stores it data,
etc, would all be great on the website. Also I'm still having trouble
understanding how Longwell 'names' nodes; I can't seem to get it to
use foaf:names, for example. There's an email indexed on google
somewhere about this, but nothing else.) It's a far more polished
program than you'd ever guess from the documentation. Perhaps one of
the dangers of programs produced in academia -- too little incentive
to hang around for the boring documentation part of the work! (I say
this from a position of considerably deeper culpability than whoever
worked on Longwell ...)

Scott
_______________________________________________
General mailing list
[email protected]
http://simile.mit.edu/mailman/listinfo/general

Reply via email to