Back to focus RE: More obscure analogies

Tue Jun 7 05:21:02 PDT 2005

Tony

> Maybe we should rename this group the 'VO Philosophy' IG? :)

Count me out ... seriously I apologize for having contributing to lead the debate astray.
For my defense : I was not alone :))

> We need to focus on what the voevent ontology will be *used* for
> and then look at the entities and relationships we need to define for that.

Agreed - the rest of this message tries to be re-focused on technical stuff

Out of all recent brainstorming, I guess there is some sort of agreement on the following
needed entities. First independently of the VOEvent data format, which is the support for
information exchange about those entities.

- "Data Packet" : Set of data gathered under some identified, homogeneous conditions,
including instrumentation, experimental protocol, team, format of data, date/location ...
Seems that each VOEvent includes at least one Data Packet (not sure if it's exactly one,
or one or more).

- "Interpretation" : Assertion of a n-ary relationship between
	One or more Data Packet
	One or more associated Event
	One or more associated Object
	One or more Hypothesis
	One or more Author

Cardinalities are open to discussion ...

Interpretation has an unique ID
More properties can be added, like "Quality", and an open list of relationships with other
interpretations such as consistentWith, inconsistentWith, supportedBy, validatedBy,
refutedBy ...

- "Event" : More or less transient phenomenum (limited interval of time), defined in a
space-time framework. Event has an unique ID and at most one asserted type. Event types
are defined and maintained in specific vocabularies by relevant authorities, or as
subclasses of "Event" (two options worth considering).
Example : GRB, SN, Solar Flare, Star Crisis ...

- "Object" : More or less permanent space-like entity
Object has an unique ID and at most one asserted type. Same remarks as for "Event".
Example : Planet, Star, Galaxy ...

Note that since "Event" and "Object" have a proper life and identity, new Interpretation
can be added at any time to the same Event, using the same or other Data Packet.

Different Interpretation can link the same Data Packet to different Events or Objects, and
associate the same Event to different Object, different Hypothesis, etc So this kind of
model will support the following kind of queries :

- Find hypothesis about an event X, matching data collected by instrument Y, using format
Z ...
- Find events of type A associated with an object O ...
- Find all Interpretation of Data Packet D, where associated object is O ...
- Find all Interpretation associated with event X, consistentWith, or inconsistentWith,
Interpretation I.

Note that for the last one, certainly the most interesting use case, the result of the
query may use declared inconsistencies or inferred ones (based on e.g. conflicting
associated object or event type).

And so on.

Now the question is to clarify which type of entity, and how many of them is/are attached
to a single VOEvent (defined here as an XML document). I would tend to think that a
VOEvent corresponds exactly to one Interpretation. That sounds too good to be true.  What
do you think?

Bernard

-----Message d'origine-----
De : owner-semantics at eso.org [mailto:owner-semantics at eso.org]De la part de Tony Linde
Envoye : mardi 7 juin 2005 11:40
A : semantics at ivoa.net
Objet : RE: More obscure analogies

Maybe we should rename this group the 'VO Philosophy' IG? :)

<<My original comments were specifically about the word "event". An event is not only an
arbitrary notion - it is a whole constellation of arbitrary notions.>>

Surely, from the sem.ig point of view, we need to focus on what the voevent ontology will
be *used* for and then look at the entities and relationships we need to define for that.
Let's keep things simple and focussed to start with and then evolve (sexually and
asexually) the ontologies as new applications or needs arise.

Do we have any use cases for a voevent ontology?

Cheers,
Tony.

From: owner-semantics at eso.org [mailto:owner-semantics at eso.org] On Behalf Of Rob Seaman
Sent: 07 June 2005 07:46
To: semantics at ivoa.net
Cc: Rob Seaman
Subject: More obscure analogies

On Jun 6, 2005, at 4:28 PM, Bernard Vatant wrote:

Cool. I love recursive definitions :) That one is certainly operational - fit for everyday
life affairs. But to provide a scientific definition of human kind as a species, forget
it, it's easily broken ...

Actually, I think it is fairly close to what would be regarded as the current "scientific
definition". More to the point, it isn't clear that "everyday" and "scientific"
definitions are, or should be, different. Certainly complexity of expression doesn't equal
depth of understanding.

I would be happy to trace my ancestors back to, say, 200 millions years ago, and meet them
one by one - just to check the humanity breaking point

But evolution contains the context for addressing these issues, too. Our most recent
common ancestor with chimpanzees was something like seven million years ago. See Richard
Dawkins' fabulous "Ancestor's Tale".

... Not to mention my descendents 200 millions years from now

It may indeed be impossible to describe entities that do not yet exist.

"When in Earth history did humanity begin?" is a question exactly similar to : "When in
Universe history did galaxies appear"?

Well, no. The underlying "equations of state", if you will, are very different. Galactic
evolution is evolution in name only. You've already pointed out the chicken and egg
problem for biological entities. Life is a rich binary tree of cousins - many branch
points, many "buds". Galaxies were always galaxies. Before that, they were only
proto-galaxies. Our non-human ancestors had their own identities - they were never
"proto-humans". One would be more correct to compare the first galaxies to the appearance
of the first eukaryotic cell, perhaps - but even then the complexity of the web of life is
both greater (in variety) and less (in the knowledge compression of evolutionary genetics)
than that of "mere" astronomy.

a species is still something very difficult to define, and hard to observe. I won't
elaborate on the details here, but I had an interesting breakfast with a botanist a while
ago, which definitely destroyed in my mind the notion of any "objective" definition of
what a species is ...

It may not be surprising that the taxonomy of Linnaeus may need tweeking after 250 years.
The notion of a species is quite different - perhaps nonexistent - for entities that
reproduce asexually. See http://ucjeps.berkeley.edu/map2.html.

"Undecidable" was not adequate a word, I agree, if you read it in its strict logical
sense. I mean you can chose any arbitrary limit, like geographs do for the classification
of cities, by setting arbitrary population thresholds ...

But "city" is an arbitrary notion. "Human" and "Galaxy" have intrinsic meaning, no matter
how difficult to pin down. My original comments were specifically about the word "event".
An event is not only an arbitrary notion - it is a whole constellation of arbitrary
notions.

Rob Seaman
NOAO