Rob Seaman wrote:
>
> My previous question about use cases sank without a trace. Perhaps I
> can try again. Here you are making a "case for using this tool", not
> asserting a use case. Before learning to use a tool, one needs to know,
> as Edwin Starr says: What is it good for?
Obviously, the tool allows for complex query. It is a graphic tool because otherwise the mind starts to boggle when trying to re-parse long queries a few weeks later.
>
> Several times during this discussion folks have indicated that such
> ontologically aware tools are useful in various other fields that are
> parsecs ahead of astronomy in this regard. Let's get past that. In
> what *specific* ways can these great tools be used to accomplish
> astronomical chores - whether big or little?
Are you looking for specific astronomical use cases? Here is one mostly for complex table data. It was: Find spiral galaxies in a Supercluster named "Hercules" that are brighter than 12th mag in I-band within the Holmberg Diameter and have 21 cm W20 rotational velocity from Arecibo, and that have Cepheids brighter than 20th mag in I with known periods and transform this info a TF distance and Cepheid distance and compare.
Or are you looking for an abstract level of understanding? One can describe in computer understandable form subclasses that are of the form
RestictedClass =
Class property (Class Property (Class Property ...
Property (Class Property ...
...
property (Class ...
...
Operation Input Class1.
Operation Output Class2.
Operation urn "Some urn to executable or wrapper to executable that
accepts OWL objects".
The user can now draw out their intentions and an application can make
it happen.
Finally, a reasoner can help you to draw up the plan backwards from the
goal. Start with your goal. The app tells you which Operations and
which input Classes lead to the goal. Select some of these. Now the
app can tell you which Operations and Classes lead to the input Classes.
If you want, set a goal (Restricted Class) and let the app try all
paths. You can make it multithreaded. Let it run over the weekend and
then stop it and see what it got done.
> Are the "domain practitioners" in other fields ever required to
> understand - or ever even read - the word "ontology"? Samuel Johnson
> and endless generations of school teachers have made the word
> "vocabulary" a familiar friend. The word "ontology" is as opaque today
> as it was to Johnson himself:
>
> ONTOLOGY: The science of the affections of being in general;
> metaphysicks.
Let's not fear a word and especially not its etymology. Geez. Ontology means you got a classification (like zoology) of Things and you have a list (formally also a hierarchy) of properties (Verb) that have an allowed domain (Subject) and range (Object). Throw in cardinality, union and intersection and that is it. Anyone who speaks can grok it.
>
> There is some reason that ontologies have been hanging fire in the VO.
> Is it simply crass unfamiliarity? Or is it perhaps that our needs
> differ from other communities? Obviously there is pent-up interest in
> resolving this issue. Folks aren't sending these messages to the
> semantics list (and certainly aren't subscribed to the list in the first
> place) to just chat - we'll have plenty of that next week. Rather, the
> word "ontology" came up at the first VOEvent workshop as some promethean
> technology that would (eventually) help us to resolve all ambiguities
> and shadings of meaning and to close the gap between harsh machine
> representations and the subtle gradations of human expression.
There is an iteresting coincidence, if you will, that the two people pushing OWL the hardest, namely Brian and me, are the only people in the NVO with deep background in derived data search and visualization. The NASA centers and observatories that make up the NVO deal with distributing images and spectra. So, the emphasis has been on: tell me what part of the sky you want an image and we can give cut it out for you. Tell me what object you want a spectra of and what resolution and we can give it to you. Yes, it is rocket science, but it is not hard rocket science. The challenge is to make optimal use of derived data. The European part of the IVO has been willing to get its feet wet, but the NVO has not. In my opinion it is a total catastrophe. Joe/Jane astronomer will see the NVO as a failure if it does not bring home derived knowledge. Finding the appropriate image or spectra was not all that hard to do before the NVO.
>
> 1) Can ontologies deliver this?
Only if we try. It will be damn hard to get the metadata into shape for
this to work. Within the US, only NED even deals with derived data at
any level. Other fields saw there was no choice. We have no choice.
I am seeing other fields with working semantic systems go to NASA and
say "you should fund us to do astronomy". "It is easier for us to adopt
to a new domain, then for the astronomers to adopt to semantics". This
argument is looking more and more correct.
> 2) Do astronomers need this?
It is true that some astronomers can get by just fine with paper and pencil.
> This reminds me of the "to STC, or not to STC" debate. Space-time
> coordinates are clearly intrinsic to the practice of astronomy. What
> can we do to resolve the same question regarding ontologies?
>
> - Rob