pcplantdb

[Bear Kaufmann <bear@ursine-design.com>]

Hello everyone,

Best holiday wishes to all.

Well, I've been out of communication for a while, if you hadn't noticed. 
Primarily school work and other work leaving me little time to work on 
the project. But now that that's over, I can spend a little more time on 
this project.

I've actually been working on some things in the last few days related 
to the project. And did install the block list, so hopefully we won't be 
getting wiki updates from spammers.

I was re-inspired to experiment with and learn the RDF structure. I did 
some work with regexs and parsing to load in data from USDA Plants,  
PFAF comments, and USDA GRIN 
(http://www.ars-grin.gov/cgi-bin/npgs/html/taxgenform.pl).
I loaded my parsing work into a table called attributes:
id           int(16)
subject      int(16)
predicate      varchar(100)
object      varchar(100)
qualifier      varchar(100)
input          binary(1)
source      varchar(6)

Subject is currently just the id of the plant, predicate is the linking 
term, and object is the linkee. qualifier further defines the term. And 
source is the data source. This really isn't a "triple-store", just for 
hacking and experimentation. But does raise some questions I have about 
how to represent information fully in RDF. Namely the qualifier issue. 
I'd like to be able to easily state an amount, or some other 
"qualification" of the link, for example: prefers very WetSoil. I 
believe this can mainly be handled by "subClassing" the connectionType 
and Object. WetSoil is a rdf:subClassOf Soil. prefers is a subClassOf 
response/'attitude'/some better neutral term.

Here's some output from the table:
Subject: Predicate: hasCommonName Object: Qualifier:
Plant (): Abelmoschus esculentus hasCommonName gumbo () from GRIN
Plant (): Abelmoschus esculentus hasCommonName kacang bendi (Mal.) from GRIN
Plant (): Allium ampeloprasum hasCommonName Sommerlauch (Ger.) from GRIN
Plant (): Allium hookeri hasCommonName kuan ye jiu (Chin.) from GRIN
Plant (): Allium kurrat hasCommonName kurrat (Arab.) from GRIN

Subject: 3 Predicate: Object: Qualifier:
Plant (3): Abelmoschus manihot cultivationComment It grows well in an 
ornamental vegetable garden () from 200
Plant (3): Abelmoschus manihot ediblityComment Young leaves - raw or 
cooked () from 183, 2
Plant (3): Abelmoschus manihot growsIn Cultivated Beds () from PFAF
Plant (3): Abelmoschus manihot growsIn Other Habitats () from PFAF
Plant (3): Abelmoschus manihot hasCommonName sunset-hibiscus () from GRIN
Plant (3): Abelmoschus manihot hasEconomicValueAs Pesticide (potential 
for disease vector control) from GRIN
Plant (3): Abelmoschus manihot hasEconomicValueAs Food (vegetable) from GRIN
Plant (3): Abelmoschus manihot hasEconomicValueAs Environ. (ornamental) 
from GRIN
Plant (3): Abelmoschus manihot hasEconomicValueAs Food (vegetable) from GRIN
Plant (3): Abelmoschus manihot hasHabit Perennial () from PFAF
Plant (3): Abelmoschus manihot hasHardynessOf 9 () from PFAF
Plant (3): Abelmoschus manihot hasHeightOf 200 (cm) from PFAF
Plant (3): Abelmoschus manihot hasWidthOf 50 (cm) from PFAF
Plant (3): Abelmoschus manihot isDistributedIn widely cult. in tropics 
() from GRIN
Plant (3): Abelmoschus manihot isDistributedIn Ind. Subcont. () from GRIN
Plant (3): Abelmoschus manihot isDistributedIn China () from GRIN
Plant (3): Abelmoschus manihot isDistributedIn Australia () from GRIN
Plant (3): Abelmoschus manihot isDistributedIn Malesia () from GRIN
Plant (3): Abelmoschus manihot isDistributedIn probably elsewhere in 
tropical Africa & Asia () from GRIN
Plant (3): Abelmoschus manihot medicinalComment The bark is said to be 
emmenagogue () from 240
Plant (3): Abelmoschus manihot pollinatedBy Insects () from PFAF
Plant (3): Abelmoschus manihot propagationComment Seed - sow March in a 
warm greenhouse. The seed should germinate with two weeks, when it is 
large en () from

This format is a pretty far cry from a real RDF store, but gives me a 
taste of the possiblities. After experimenting with this, I wanted to 
explore RDF more fully. One of the RDF query languages I was looking at 
called RDQL is rather similar to SQL.
SELECT FROM * (namespaces)
WHERE (?x, <subClassOf>, ?z)
AND ?x=condition
FOR namespace=url, piw=url, dc=url....

The WHERE clause is the core of it, and some of the beauty of RDF. If we 
wanted to find all the outgoing links from a plant we could set WHERE to 
(<piw:Somethycum_bicolor>,?y,?z). With that query we'd get basicly all 
we'd need to show a plant.
WHERE (?x,?y,<piw:Somethycum_bicolor>) would tell us what links to it. 
Perhaps goats consume plant. etc. BTW, there can be more then one part 
in the where statement which can allow deeper searching, but it seems 
(at least with the implementation I'm using, which may be handling a 
good deal of checks within PHP) that complex "deeper" queries require a 
good deal more computing time (understandably, since it's doing JOINs 
and UNIONs and such). So I'm inclined to try to define plants and other 
objects as much as possible going only one level deep with the links. 
It's a call between semantic richness and computational effiecency.

OK. That's enough for one letter. I'll write more about my thoughts 
soon. I've been working on moving Rich's XML schema doc towards an RDF 
Ontology. While this letter is rather technical, I think there should be 
some interesting discussion on how we could lay out vocabulary 
(Ontology) that will fit the needs of the DB.  I think this avenue is 
very interesting in terms of the flexiblity it provides in defining the 
schema of plants and other objects (and interface too me thinks). Sorry 
if I didn't do any softer leadups to any terms I used. If anything is 
confusing, let me know.

Once again, hope your all well.

Talk to you soon,
Bear