[QUAR] Re: Expressing position in RDF

Tue Oct 14 14:39:20 PDT 2008

Matthew

Let me be practical if I can :-)
> I'm working on the data portfolios for VOEvents that some of you have 
> heard me talk about (at Practical Semantic Astronomy 2008 and IVOA 
> Trieste). Each VOEvent gets associated/translated into a XHTML/RDFa 
> document that then gets added to as more data/information about the 
> event becomes available. In this particular instance, I have the STC 
> from the VOEvent which will generate an XHTML statement: "This object 
> has RA= ... and Dec=..." but will also have the same information 
> expressed in RDFa.
>
> The portfolios get parsed and the RDF(a) put in a triple store or data 
> discovery launched because some criteria are met, etc.
The point of publishing data in RDF is the "open world assumption". To 
answer question b) of Douglas below, the answer of RDF is "you never 
know". Actually you put your data in RDF in order to allow any kind of 
future merging with data you don't know of. In the closed world of 
VOEvent your "RA" property will not be ambiguous maybe, but will it 
merge properly with other "RA" properties?
If you don't have such questions in mind, RDF is useless, actually.

Bernard
>
>     Cheers,
>
>     Matthew
>
> On Oct 14, 2008, at 12:15 PM, Douglas Burke wrote:
>
>>
>> Matthew,
>>
>> Perhaps if you can place your request in a broader context - I'm 
>> thinking along the lines of
>>
>>  a) what system is going to create these statements about coordinates
>>  b) what system is going to use them
>>
>> - so that we might begin to agree on what is simple for this 
>> particular task. Actually, I doubt we'd get that far, but I'd be 
>> interested to know what you're actually trying to do by 
>> "sematicising" positions ;-)
>>
>> Doug
>>
>>
>> Matthew Graham wrote:
>>> Hi,
>>> I'm happy for the philosophical discussion but am also trying to 
>>> figure out how to actually do something empirical. I can see exactly 
>>> the same arguments that we had about using STC in VOEvent applying 
>>> to representing positions in RDF. The 90:10 rule should apply and 
>>> whilst it is wonderful that I can describe any position in any 
>>> coordinate system using ontology X, why can't we have ontology Y 
>>> that is small and simple (that word again):
>>> :myStar stc:UTC-TOPO-FK5#RA 134.56
>>>    Cheers,
>>>    Matthew
>>> On Oct 14, 2008, at 11:47 AM, Rob Seaman wrote:
>>>> Bernard Vatant wrote:
>>>>
>>>>> how do you attach measurement results to an object?
>>>>
>>>> This is begging the question in an observational science.  The 
>>>> existence of the object is the null hypothesis you're trying to 
>>>> test.  Anyway, I don't think Matthew was looking for a 
>>>> philosophical discussion :-)
>>>>
>>>> But if you did want to pursue this, one might suggest starting with 
>>>> the distinction between the dependent and independent variables of 
>>>> the observation/measurement.
>>>>
>>>>> Seems to me the scientific community (at least its members 
>>>>> involved in Semantic Web) should try and standardize this at some 
>>>>> point.
>>>>
>>>> This is the center of the scientific maelstrom.  One might have 
>>>> better luck first looking elsewhere.
>>>>
>>>>> I've been looking for relevant pointers to people working on this 
>>>>> in other domains (say e.g., Biology, Earth sciences ...) without 
>>>>> much success so far I'm afraid.
>>>>
>>>> Returning to the question at hand, this is an interesting point.  
>>>> We act as if astronomical coordinates are particularly difficult.  
>>>> They are, in fact, particularly well behaved.  Things tend to stay 
>>>> put on the celestial sphere (for many purposes).  Imagine building 
>>>> the same assertions for - say - wildlife management.  A herd of 
>>>> caribou doesn't stand still.
>>>>
>>>> One may, however, make assertions - as with stars - about the 
>>>> temperature of a particular caribou or other parameters such as sex 
>>>> or mass or age.  These are assertions inherent in the object 
>>>> itself.  One may make assertions about group behavior - a "cluster" 
>>>> of caribou.  One may make assertions about evolutionary descent.
>>>>
>>>> But it is orders of magnitude more difficult to specify location 
>>>> (longitude and latitude as a function of time) for caribou than it 
>>>> is for stars.  And in astronomy, one is typically expressing 
>>>> targeting coordinates (explictly or implicitly) for future 
>>>> observations.  On the other hand, predicting the future migrations 
>>>> of caribou is simply impossible.
>>>>
>>>> It seems unremarkable to me that an assertion in an astronomical 
>>>> context (say, stars), might look something like:
>>>>
>>>>    X is a star
>>>>    X corresponds to target Y
>>>>    Y has WCS Z
>>>>    Z has an RA (along with a bunch of other attributes) - and 
>>>> corresponding to some fiducial point like the centroid of a PSF
>>>>
>>>> Compare to:
>>>>
>>>>    A is a caribou
>>>>    A is somewhere in Alaska
>>>>    Alaska has (complex and idiosyncratic) GIS data structure B
>>>>    B has a footprint the size of the lower 48 east of the Mississippi
>>>>
>>>> Rob
>>>>
>>
>>
>
>

-- 

*Bernard Vatant
*Senior Consultant
Vocabulary & Data Engineering
Tel:       +33 (0) 971 488 459
Mail:     bernard.vatant at mondeca.com <mailto:bernard.vatant at mondeca.com>
----------------------------------------------------
*Mondeca**
*3, cité Nollez 75018 Paris France
Web:    www.mondeca.com <http://www.mondeca.com>
Blog:    Leçons de Choses <http://mondeca.wordpress.com/>
----------------------------------------------------**