Re: Astronomy meets Semantic Web/Linked Data?

Greetings, all.

I can add a few things to Kjetil's message, none of them disagreeing 
with his essential point.

On 3 Apr 2016, at 23:37, Kjetil Kjernsmo wrote:

>> It seems our cousins in Astronomy have lots of problems coping with
>> large amounts of diverse data, and we have the typical integration
>> problems across different observatories, as well as questions of how 
>> to
>> make data public in a reusable manner, and so forth. So trying to 
>> apply
>> SW/LD methodologies to the area of Astronomy would seem to make a lot 
>> of
>> sense.
>
> Yes, indeed it does. There are many problems, for example 
> multi-wavelength
> studies, transient objects, large surveys, classifications, automated
> hypothesis generation, where Semantic Web technologies make an awful 
> lot
> of sense.

This is very true.  Astronomy is a highly-collaborative culture, where 
people routinely share data between groups and between continents.  The 
relatively heterogeneous data is often (by design) highly compatible, 
but there are many cases -- mostly the interesting cases -- where there 
are subtle differences in meaning which must be preserved.  Astronomers 
care about the provenance of their data, and care about preserving it, 
and its meanings, for the long term.

It sounds like a natural application of SemWeb technologies, and at some 
level it is.

> In Europe, Norman Gray has been the main champion of Semantic 
> Astronomy.

That's kind of you, Kjetil.  I did do a lot of banging of this 
particular drum, but there were other true believers, including people 
at Strasbourg (Sébastien Derriere) and Caltech (Matthew Graham).

Ismael and Aidan: you may want to look at the 'Semantics' documents at 
<http://www.ivoa.net/documents/>; the IVOA Semantics WG pages linked 
from there are now slightly out of date.  Although I don't think it's 
listed on that page, for some reason, the most hard-core SemWeb ontology 
that appeared from this consortium is 
<http://www.ivoa.net/documents/latest/AstrObjectOntology.html>, which is 
an OWL ontology for astronomical objects, designed with the intention 
that it be usable for drawing relatively high-level conclusions from 
data; that was a very nice piece of work, but got very little traction.

I ended up believing that the most effective way to introduce these 
technologies into astronomy was through a much more lightweight 
approach, based on linked data and vocabularies, rather than full-blown 
ontologies.

> However, it seems we have failed to impress astronomers. Astronomers 
> tend
> to work on pretty heavy things, and many are themselves capable
> programmers, but many are also skeptical about the hypes of the 
> industry.
> Semantic Web has been difficult to sell as more than just another 
> bandwagon,
> I suppose. They are keen to use academic software, but it has to be 
> rather
> mature, or developed by astronomers :-)

What Kjetil said....

I could possibly write quite a lot about why SemWeb paradigms failed to 
get any traction in this community, and perhaps I will, some day, but 
the highlights are below. (by the way, I still call myself an 
astronomer, despite the fact that almost all the 'astronomy' I do is 
computing-based; I was several times accused of being a computer 
scientist, and not in a nice way...).

Astronomers have been sharing data electronically since the 70s, both on 
tape and online, and the still-current astronomy transport and archival 
standard, FITS <https://en.wikipedia.org/wiki/FITS> was first 
standardised, based on widespread practice, in 1981.  Astronomy data, 
rather like particle physics data, has always been at the technological 
limit of what can be feasibly managed and transported.  Astronomers are 
therefore entitled to believe that they have a pretty good handle on how 
to 'do' data.  They're as interested in shiny toys as anyone else, but 
there are plenty of folk in the community who have spent their entire 
professional careers watching computing fads go by, and who therefore 
take a certain amount of persuasion that technology X has something 
genuinely valuable to offer.

Everyone can code.  Even full professors can generally write at least 
some code in, say, python, that they might be willing to share.  That 
is, 'here is some software magic! close your eyes and Believe!' is not 
going to work in this community in the way it might in, say, biology.

Everyone can code, but no-one is interested in computers per se, so 
no-one is excited by software novelties, or interesting new computing 
paradigms, unless they demonstrably process data faster or better or in 
some way that wasn't feasible before.  There _is_ a cadre of folk who 
_are_ generally interested in computers and who have made their 
scientific contributions that way, and have become (not sure of the 
word) 'astronomer-technologists' -- a subset of the people in the IVOA 
(including me) are of this type, and there's a journal where (quick 
plug!) we can publish 
<http://www.journals.elsevier.com/astronomy-and-computing/>.  But they 
(we) are to some extent 'dropouts', and are not really in a position to 
lead the discipline.

One could characterise this as 'conservatism' (and I have, loudly and 
crossly, in the past), but it's really more pragmatism.  I/we failed to 
make the case, in astronomy, that the SW paradigm did something that 
genuinely couldn't be done otherwise.

Because FITS works.  FITS as a data format, and as an intellectual 
framework, has many defects, at multiple levels (see special section at 
[1]), but it does do the job well enough in astronomy that there isn't a 
problem _which is blocking progress_, which SemWeb can solve.  The 
blocking data problems in astronomy tend to involve volume rather than 
semantics.

Geoinformatics is slightly different.  Geosciences has some intellectual 
overlap with astronomy (solar physics, for example, is 'astronomy' in 
Europe but 'geophysics' in the US); it has very heterogeneous data, 
often in smaller volumes than astronomy; it has a community which 
doesn't always have the software nous to force through its own 
solutions.  If you're interested in applying SemWeb ideas in this 
general area, then this might be worth looking at.  I don't have a 
particular reference for you, but if you google 'geoinformatics peter 
fox' you should be led to interesting things.

So, Aidan and Ismael, I think there are good things to do here, in an 
interesting technological landscape.  If you're interested in more (even 
more!) detail, then feel free to contact me off-list.

All the best,

Norman


]1] http://www.sciencedirect.com/science/journal/22131337/12


-- 
Norman Gray  :  https://nxg.me.uk
SUPA School of Physics and Astronomy, University of Glasgow, UK

Received on Tuesday, 5 April 2016 12:54:58 UTC