Data Portal discussion from techwg

[Ed Shaya]

Toward the end of a lengthy e-mail at techwg, Tom McGlynn suggests that 
the next thing to do is to develop a format for DAL requests.  One might 
imagine that coming to agreement on such a language would be a long 
arduous task (as in years and years) and, in the end, one would have 
something that still did not do even a majority of what users want. 
However, Ontology already exists and it can provide such a request 
language.  This is because queries can be viewed as constraining the set 
of everyThing down to the set of things that you want.  So, you just 
need to make triplet statements that do this constraining.  Not only 
does the language exist but the necessary reasoning engines exist as 
well.  Let's not reinvent the wheel [VOwheel]. For the examples given by 
Tom, I show how this might go.  Of course, with the rich language that 
this requires the user would need an application that guides them by 
selecting off of menus until they become fluent.

> 1. Define a  file format that represents a sequence of DAL requests. 
> This format would be trivial for really simple requests, but would 
> support requests where later steps use as inputs the output of earlier 
> steps.
> 
> Here are some examples of what I think users would want to be able
> to do (please don't get hung up on the format)..
> 
>    a. A simple cone request might just be
> 
>         Query name=chandra position=3c273

   We would assume that the following is already in the system.
3c273 isa Quasar
Chandra isa Telescope

   Now we request a List of Images with ?Pos in Image Area.  [The
   ontology is all in the singular, even tho plural is implied.]
3c273 hasPosition ?Pos
?MyList isa Image
?MyList takenWith Chandra
?MyList hasFOV ?FOV
?Pos in ?FOV


>    b. Look for all of the long exposure Chandra images that are match 
> a set of position from a file.
> 
>         Req1: ImageQuery name=chandra size=0.1 position=@mylist.file
>         Req2: Filter Req1 exposure>10
>               Retrieve Req2

   First make NewList with Images with FOVs including MyList positions.
MyList isa Position
?NewList isa Image
?NewList takenWith Chandra
?NewList hasFOV ?FOV
MyList in ?FOV

	We add exposure time filtering
?NewList exposureTime Time1
Time1 hasValue Range1
Range1 greaterThan 10
Time1 hasUnits hours

> Here 'Filter' applies a filter to a VOTable and Retrieve retrieves 
> data references in an appropriate VOTable.
> 
>    c.  Not sure if the chandra SIAP service is called Chandra?
> 
>         Req1: ServicesQuery keywords=SIAP,Chandra
>         Req2: ImageQuery  Req1 size=0.1 position=@mylist.file
>         Req3: Filter Req2 exposure>10
>         Retrieve Req3
> 
>    d.  Retrieve a list of targets from  a list of high-res optial 
> resources.
> 
>        Query name=sdss,usnob,gsc2.3,2mass-j,2mass-h size=6"
> 		position=@mylist.data
>        ImageQuery name=2mass-k,hst position=@mylist.data

     This is just asking for a cross-correlation, but we don't need thE
	VO for that.  Below I do the more general task of requesting the
	info directly from datacenters:

       Query Resources SDSS,USNOBtable,GSC2.3table,2MASS [This is
	directive to Application, not really in ontology)

	 Set up Objects to return
?S isa AstroObject
?S hasPosition ?P
?S hasMeasurement ?m

      Add 6" error to positions in MyList
MyList isa RADECValue1
Position hasValue RADECValue1
Position hasAccuracy Acc
Acc hasValue 6
Acc hasUnits arcseconds

?m union ?m_J,?m_K,?m_H,?m_V,?m_R  (presume O and E images have been
	  translated to V and R)
?m_J inBand J-Band
?m_K inBand K-Band
?m_H inBand H-Band
?m_V inBand V-Band
?m_R inBand R-Band
	[We could instead have: (?M inBand ?Band) and let data center 			fill 
this in for every band they have.  Depends on whether you 			want to 
control output exactly.]
?P in Position


I found it odd that none of Tom's examples needed any astronomical 
functions or transformation, but those can be embedded as well.

So, how would this all work?  These three word sentences are easily 
converted to OWL.  Then we need OWL to various query tools, VOQL, SIAP 
etc.  We have been working with OWL to XQuery and getting the ADC tables 
to respond to XQuery.

Ed
-------------- next part --------------
A non-text attachment was scrubbed...
Name: eshaya.vcf
Type: text/x-vcard
Size: 257 bytes
Desc: not available
URL: <http://www.ivoa.net/pipermail/semantics/attachments/20070710/f211ff6e/attachment-0001.vcf>