Fwd: some remarks on VOEvent

[Ed Shaya]

Rob Seaman wrote:

> On Jun 4, 2005, at 7:13 AM, Ed Shaya wrote:
>
>>   I'm afraid I don't get what you are flaming about here.  Why is it 
>> possible to define a VOEvent but not an event?
>
>
> My apologies if I seemed to be flaming.  I thought we were just having 
> a conversation - expressing diverse views.
>
> A VOEvent - as Roy helpfully pointed out - is the report of an 
> observation.  Hundreds of years of scientific developments have 
> provided very fine grained utilitarian definitions of "observation" 
> and "measurement" and other experimental concepts - and of specific 
> flavor reports of same.  "Event" is a nice flexible English word - but 
> more on the theoretical side than the experimental.  In its colloquial 
> sense, the word "event" is likely broader in its connotations in 
> astronomy than in any other discipline.  I thought the point of 
> ontology was precisely to capture colloquial meanings in a bottle.  I 
> suspect "event" is one such word that will be hard to so capture.  

There is no bottle requirement.  Where one can define precisely, one 
does.  Where one can't, one does not.  The ontology will not collapse if 
the words big and little are in there.
Also, ast:event does not need to be the same as top_ont:event.  If you 
want, there can be a vo:event that is separate from ast:event.

> My worry - a worry that I've heard others express without deeming them 
> to be flaming - is that "ontology" is a synonym for "non-pragmatic".
>
We of this workgroup do hereby deem this as flaming and ban such 
expressions from this list forevermore.  All agreed, say aye.

>> Ontology allows one to use natural language to make statements and 
>> query.  The whole idea is to bring the flexibility of human language 
>> to machines (but perhaps to make it a bit clearer).  If you think 
>> this is a vain hope then why are you on this list?
>
>
> I don't think it is a vain hope to build targeted ontologies - note 
> the plural. 

Look at the VOEvent.owl.  It has no superclasses and no subclasses.  The 
properties are all hasObject where Object is simply the thing it has, 
every time.  This is what happens when you create targeted Ontologies 
that are small and not tied into overarching ontologies.   If we are to 
have a consistent vocabulary and more importantly if we are to make 
progress in our lifetimes without getting bogged down by the gridlock of 
interdependencies we need overarching ontologies first.  Afterwards we 
can go deep into specific subfields. If you look at the ontologies I've 
been working on there are targeted ontologies: physics, geometry, 
chemistry, astronomy, general science, and quantities.  Not the targets 
you had in mind.  But all quite necessary.

> 'm on the list because all of a sudden there are four or more lists 
> for discussing VO semantics and we keep being shuffled off to the next 
> list over.  VOEvent presents an interesting opportunity for the VO to 
> wrestle with real world semantic requirements.  Should this particular 
> list be renamed "ontology at ivoa.net <mailto:ontology at ivoa.net>"?

Maybe so.  It was resurrected because I was requesting an [Ontology] 
subgroup in DM.

> Can semantics be profitably pursued as an activity separate from other 
> VO work?
>
I think the idea is for this group to experiment with Ontology and when 
the benefits to the other groups are clear we will inform them.  If this 
is successful then the day may come when UCD and DM are superfluous. 
Also, there would be many uses for any good scientific ontology, so this 
work may be profitable to a much wider group than the IVOA.

>> Ah.  So what you want is inflexibility.  You want to ensure that 
>> VOEvents do not have explanations of explanations or any other deep 
>> discussion of the event.
>
>
> Not what I want at all.  Although I might suggest that "deep 
> discussion" is an emergent property from a long sequence of richly 
> braided followup packets.  To optimize its utility, a single VOEvent 
> will typically express a single coherent observation or hypothesis.  
> VOEvent is a specific VO facility with a clear mission - to report 
> alerts to the community via software agents.  In many instances, 
> VOEvents will be used to trigger more in depth analysis via other VO 
> facilities.  An "alert" and a "deep discussion" are indeed quite 
> different modes of communication.

>
>> If the VO wants hardwired structures it should study the issue with 
>> Ontologies (because only in ontologies can you see the big picture 
>> with relationships between all of the components of the langauge)  
>> and then use that as a starting point for creating limiting UML or 
>> Schema.
>
>
> "Limiting UML"?  :-)  If an ontology is something so open-ended that 
> UML is seen to be limited by comparison, then perhaps Merriam-Webster 
> is right to call ontology a branch of metaphysics.  I was just 
> standing up for ontologies the other day on one of the gazillions of 
> UCD lists.  There is little point to UCDs without an underlying 
> ontological formalism.  But by the same token, there is little point 
> to ontologies if they aren't seen to be pragmatically wrestling with 
> real world issues - at least on occasion.

> VOEvent is an opportunity for elucidating real world requirements and 
> building real world VO systems.
>
UML and Schema are limiting in the sense of providing a predefined 
structure and order.  In RDF the information can come in any order and 
the root is always RDF so one can add any desired extra information.  It 
is not clear which is more pragmatic, but if you happen to need 
structure then by all means create a schema.  You can annotate the 
schema with terms from an ontology.

> Phenomenologically yours,
>
> Rob
>
>