[openspending-dev] Openspending - one year later Was: Working with OpenSpending platform from my perspective

[Friedrich Lindenberg]

Hey Peio,

On Tue, Jul 10, 2012 at 5:59 AM, Peio Popov <peio at peio.org> wrote:
> One small detail, that I notice, is that in Bulgarian the currency sign is
> behind the number - it should be 100 лв. and not лв.100. Another thing is
> that 502 error I am getting when trying to manage the dataset, but it is
> something temporary I guess. So far - nothing that stops me for using it or
> makes it somehow not pleasurable.

Thanks for these two. The currency thing is easy to fix, we basically
have a mapping for this for all currencies (but are still propagating
the use of that library through OS). The timeout errors come from
apache solr, which for some reason has decided that it wants to be
massively unstable. I'm investigating the issue, but to be honest
don't have a trace yet.

> On a related note - Rufus mentioned you were getting into linked data. What
> I would like to go next is a simple project to link the data from the daily
> spending with the data for the minister/official responsible for it and
> quantify his importance in such manner. Something like who spends our money
> or who is responsible for the largest spending in the country and give a
> idea about the monetary power of the person. Do you have any spending
> ontologies and/or ideas for such projects?

It'd be very interesting to try more data linkage - as you describe to
responsible politicians or output/impact metrics (how do school grades
react to a cut in education spending?). I think the way to do this is
a bit painful: manually checked reconciliation is the best we have for
most data. I'm very keen to do a web-based matching server (where you
can upload a reference list and then compare and match individual
entries), simply because I haven't come across one yet. This must
exist somewhere.

Converting stuff to RDF just introduces a large degree of complexity,
theory-heavy processes and a sucky tool chain (while making it so much
harder to use existing tools for cleansing and visualization) without
actually tackling the actual problem: how do we *actually* do record
linkage? Using URIs is a side concern, and I'm fully on board with it
- but the problem most of the time is about linking things like
department names to global identifiers. That is hard with or without
linked data. In either case, I think the policy for OKF is to stay
well clear of all things RDF except where we're required to do
otherwise.

Cheers,

 - Friedrich