[pcplantdb] RDF and Wikis
bear at ursine-design.com
Thu Jan 12 01:54:11 EST 2006
John Schinnerer wrote:
>Did, briefly...couldn't make heads or tails from what was on those web
>pages. However your description of intent and application below is very
Yes. I was heading out and provided little explanation of what I did.
If you want to explore it, try looking at Classes. Then say look at
Plant. Unfortunately this documentation system "flattens" the whole
thing, but you can see the parents of the classes (Superclasses) when
you view a class. Or try individuals and look at Cucurbita_moschata, one
of a few plants where I tried to describe it with the schema, and
develop what was lacking.
I've appended this letter with the current subclasses of the ontology.
It's not fully organized, and would probably end up with parts falling
into very broad categories (States, Conditions, Organisms, Construction,
GeographicArea, etc). I worked from the Eden schema, as well as USDA
Plants, started to add soil classifications from a USDA Soil
Classification workbook, etc.
>I like it.
>As far as I can make it out at a quick read, you pretty clearly indicate
>how it would support multiple goals in our project.
>One of those goals is the IMO very important one of providing an
>extensible framework for "guiding" user input in as structured of ways
>as we find necessary.
Yes. Matter of fact I was envisioning something like Google Suggest (not
to jump on the AJAX bandwagon, :-), which could guess what someone was
typing, and suggest terms in the ontology. There's an OWL property
called "sameAs", so terms that were the same, but where we use a
different term, could guide the user saying: "We use this term x instead
of y, do you feel x conveys what you mean well enough? Otherwise, check
this box to submit a suggestion for extensions to the schema." Similarly
unique terms could get placed into a queue to be incorporated into the
schema, or be assigned a "sameAs" property.
>For what afaik we have said we are trying to do, a wiki is clearly not
>at all even a little bit suitable. I thought we'd been through all that
>already long ago.
>Am a bit surprised to see wikis mentioned again as anything other than
>accessory/supporting to the actual plant DB.
>There's always a tradeoff between "freedom" and "structure" and wikis
>are way too "free" for what I have understood we intend to offer.
Yes. While working on the ontology, I realized that it can be very hard
to convey some things in the detail one might like, often wanting to
grab a qualifier to stick in there. So I think some form of free text
entry (which might be more natural to people, then clicking through 100
item form), but using a structured tagging using a flexible/growing
ontology behind it which can be accessed by machine, would work
reasonably. One issue to deal with in that case is how to handle data
parsed from other sources, since they'd be coming in as text free
triples, say from USDA Plants, etc.
A nice thing about a relatively small ontology composed of basically
single or double words, is that translation would be quite quick. Each
of the classes/individuals would best have a "term" property, the
English version would often just split words with spaces for nicer text
output, but someone could translate Construction, Building... into 'de',
'es', etc, etc.
Here's the class list. Under these live the individuals. Deciding what's
a class or individual is a bit difficult, but similar to OO programming:
Classes can have allowed properties defined, whereas the individuals use
those properties to define themselves.
More information about the pcplantdb