RE: The Two Dogs Problem

On 2010-12-19, Michael Schneider wrote:

>>> Since RDF is a set, the duplicate triple is disgarded.  So Alice 
>>> still is described as owning one dog.
>>
>> Alice owns _1: which is a dog, alice owns _2: which is a dog, which 
>> are not necessarily the same unless by some other logic we can 
>> conflate them.
>
> One has to distinguish between the syntactic and the semantic 
> situation. [...]

So basically you're talking about the fact that nowadays we compare 
graphs modulo leanification. I always forget about that, because it 
wasn't always so. I do that because from my relational background, it 
would be *so* much more complicated in full than from RDF's simplified, 
graded, binary model.

But do remember, even here you have to assume that the two graphs you 
gave would have to be the full graphs. If say even a single triple of 
the form _1: :hasColour :blue was added, the leanified version would be 
different. That is to say, when you look at leanified graphs only, they 
never stand alone; you have to give the whole context before you can 
infer upon them.

I'll have to defer to the real mathematical logicians around here, like 
Pat, but I do think this is an instance of nonmonotonicity, introduced 
by leanification/modulo-graph-isomorphism-thinking. Very much a 
hindrance to efficient inference, then, and probably something which has 
to do with the multiple layers of OWL we now have. (I've also pretty 
much killed myself trying to explain the precise same complexity 
boundary to my relational coworkers: how do you write up code which 
actually continuously and incrementally graph-minimizes an arbitrary, 
n-ary relational database? Don't make any mistakes, it's even harder in 
the RDF environment, in the general case.)
-- 
Sampo Syreeni, aka decoy - decoy@iki.fi, http://decoy.iki.fi/front
+358-50-5756111, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2

Received on Monday, 20 December 2010 18:42:18 UTC