Taxonomy issues

[Anita Richards]

On Fri, 27 Sep 2002, Tony Linde wrote:

> A letter in Nature today (Vol 419, 26 Sept, p 337) raises the concern
> that creating a 'taxonomy portal' (taxonomy in the sense of species
> classification) could lead to the imposition of a standard view ....

> Is it possible using DAML+OIL or equivalents to classify the same object
> as two apparently contradictory things? I assume not. But both authors
> must be allowed to classify an object as different things.... 

Is a matter of how we restrict the classification classes within an
ontology?
There are some multiple definitions which are not due to disagreements
anyway, and there is also the important point which older astronomers tend
to be more careful about, which is that we don't really know what anything
is anyway (outside the solar system places where probes have been), so
really everything should be 'has the appearance of'/'behaves like' or '
is in the direction of'
rather than 'is a' or 'is at' 
but most of us are more definate just because it is shorter....

but for example, a galaxy can have multiple names (NGC 6xxx = Markarian
348 = ...), it can a starburst galaxy and a ULIRG (ultra-luminous IR
galaxy) and a megamaser galaxy (producing bright line emission) and a
Seyfert 1 and a spiral....

So we will have to allow an object to be multiply classified.

There are a few classifications which one might naively want to make
exclusive e.g. elliptical or spiral, Seyfert 1 or Seyfert 2 (or more
complex sub-groups). However the problem you raise suggests this would
be a bad idea unless there is a very good reason for doing so.  And I think we
can have multiple descriptions?  e.g. a car can be both 'red' and 'sports'
in the example the Manchester Computing people showed us.

The trickier question is how we respond to queries, e.g. suppose one
catalogue classifies Gridlock A (the brightest radio source in a new
constellation named after AstroGrid) as an elliptical, another classifies
it as a spiral and the automatic classification routine which we run on
Sky Survey plates says it is an irregular.

If someone says 'what is Gridlock A' we can give them all 3 (or more
definitions) with references.

But if someone says 'give me all the ellipticals in Gridlock A' what do we
do?  (actually, this is the sort of thing which someone  raised at
TIVO, if you look for outliers, for every one interesting and unique
object you may get 10 or 100 misclassifications/measurement errors..)

Do you think it would work to have a 'security of classification flag'?
This would have to involve some sort of check-list of contradictory
classifications - a 'contrdiction exists' property?

This could work two ways:
When a catalogue is contributed, check objects with these classifications
against existing entries and flag those with contradictions.  Of course,
this does not mean that an uncontradicted classification is correct...

Or do the search at the time of query - for each object found classified
as an elliptical, check it is not also classified as a spiral.

Either way initially the VO could just give a contradiction flag and a
reference.  Eventually we could in some cases give an estimate of
reliability based on e.g. catalogue resolution, sensitivity, age, or even
some VO panel of experts for very large or significant catalogues (in the 
way that at present we would have a heirachy of position accuracy
catalogues, starting with Hiparchos and finishing with the Almagest (or
maybe gamma ray catalogues...).

(For stars it is worse, as they should not be e.g. both spectral type A
and G at the same time, and if they are that implies either a mistake or a
binary.  But in the case of measurements taken many years apart it could
mean it has evolved.  Galaxies don't evolve that fast (but they could have
companions within the field of view I suppose...).

cheers
a
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 
Dr. Anita M. S. Richards, AVO Astronomer
MERLIN/VLBI National Facility, University of Manchester, 
Jodrell Bank Observatory, Macclesfield, Cheshire SK11 9DL, U.K.           
tel +44 (0)1477 572683 (direct); 571321 (switchboard); 571618 (fax).