some remarks on VOEvent

[Roy Williams]

On Jun 3, 2005, at 4:57 PM, Rob Seaman wrote:

> On Jun 3, 2005, at 11:05 AM, Ed Shaya wrote:
>
>> Hypothesis - One of several possible testable explanations
>> Theory - Only known explanation that fits the observed  data.
>> Postulate - statements of assumptions needed to continue the line of 
>> reasoning.
>> Axiom - a truth commonly accepted and in no need of proof.
>> Fact -  Real world facts (New York is in the US) and Mathematically 
>> or logically proven statements.
>
> I'm concerned at an apparent emphasis here toward a "science fair" 
> model of the scientific method.  Successful science is more often 
> inductive, not deductive.  The difference between a fact and a 
> hypothesis is one of degree, not absolute confirmation via theory.  
> Theories are often only approximately correct - and approximation is 
> often much more than good enough a foundation for drawing further 
> inferences.

We model this in VOEvent by allowing the publisher of the event to 
separate information into different areas. It is the responsibility of 
the publisher to put statements in these areas.
Who (is the publisher et al),
What (are the parameter values),
WhereWhen (the space-time coordinates),
How (the observation is made), and
Why (opinions of publisher: Interpretation, Hypothesis, Classification, 
Importance, etc etc).
>
>> The VOEvent is different from  ast:Event because an instance of an 
>> ast:Event is the event itself not claims about the event.   However, 
>> times, locations and measurements of the event do belong in 
>> ast:Event. Once there is a theory of the event it can be subclassed 
>> to the specific type of event that it is (ast:supernova etc).
>
> Which begs the question of the definition of "event".  We may know one 
> when we see one - Tycho's SN was apparently hard to miss - but even 
> the distinction between an "object" and a "process" is fuzzy.  A star 
> can be viewed as a 10 billion year nuclear process.  Is an event 
> simply a process with a big amplitude of short fuse?  Define "short".  
> Define "big".

More properly we should *not* say VOEvent represents an astrophysical 
event, but rather it represents an *observation* of an event. Each 
observation is assigned an ID. Suppose we assume that when a VOEvent 
cites another, then it is concerning the same event. Therefore we can 
group all the observations that are linked through a chain of citation, 
and try to associate the group with an astrophysical source. Note that 
this group may include null observations -- "we measured flux from five 
galaxies in the error circle, and one was anomalous". Thus our 
technology needs to find these clusters of VOEvents even when they are 
published to different registries. It is the same as the distributed 
cross-match problem, but using event IDs in place of sky position.
>
> I understand a VOEvent - I've had a hand in its definition.  I fear 
> the entire astronomical community will never understand an ast:Event 
> well enough to make it useful.  (Would be delighted to be proven 
> wrong...)
>
> To put it another way, the most interesting events are like UFOs - 
> theories are hard to come by.  The benefit of ontologies and UCDs and 
> other semantic "technology" is precisely to help us to winnow out the 
> pie plates and weather balloons.  It is the UFO sightings that are 
> left after all the current theories are exhausted that we are 
> interested in.
>
> The only reason to classify Snarks is to find Boojums.