Discussion : DataLink semantics vocabulary enhacement proposal = new Associated-DataSet branch

François Bonnarel francois.bonnarel at astro.unistra.fr
Tue Oct 15 17:47:34 CEST 2019


Hi again,

On this proposal the discussion alraedy started.

For two reasons :

    1 ) some of this (associated_data) had been proposed in the previous 
page semantics DataLink Update page : 
https://wiki.ivoa.net/twiki/bin/view/IVOA/UpdateDatalinkTerms

  -- PatrickDowler 
<https://wiki.ivoa.net/twiki/bin/view/IVOA/PatrickDowler> - 2019-05-08 
wrote :

If the type of the pointed dataset is expressed this way, this semantics 
field will help the client designer to send the link for the target 
dataset to an appropriate tool , via samp for instance.

All links in DataLink 
<https://wiki.ivoa.net/twiki/bin/view/IVOA/DataLink> that are not #this 
are implicitly "associated". The proposed structure here seems to be 
drifting away from vocabulary into something else. Finally, the idea 
seems to sanction using terms from the ObsCore 
<https://wiki.ivoa.net/twiki/bin/view/IVOA/ObsCore> standard but not 
very explicitly or in any machine-readable way (that could be partially 
improved by defining an ObsCore 
<https://wiki.ivoa.net/twiki/bin/view/IVOA/ObsCore>/dataproduct_type 
vocabulary). There is a lot here that needs discussion.

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
-- AdaNebot <https://wiki.ivoa.net/twiki/bin/view/IVOA/AdaNebot> - 
2019-10-08 wrote :

For example 2: GAIA DR2 source : links to associated TimeSeries 
<https://wiki.ivoa.net/twiki/bin/view/IVOA/TimeSeries> with semantics 
field "associated_data:timeseries" -- I don't think this reflects enough 
the type of data, it is timeseries, but of what? photometry 
(light-curves), but can also be a time series of spectra or timeseries 
of radial velocities in the case of Gaia DR3. This means that there is a 
need of specialisation of the term timeseries. Same case is true for the 
case of missions providing timeseries of images

2 ) some discussion happened on the list last week following ada's 
renewal of the proposal. see below

cheers
François

Pat  2019-10-11 wrote to Ada:
> The semantics column in DataLink is supposed to clarify the 
> relationship of the linked item with the thing with the specified ID 
> -- so it doesn't necessarily tell you what the resource *is* but 
> rather how it is related.  There are cases where it looks like 
> semantics says what something is: preview sounds like it is saying 
> "this is a preview" but it is really saying "this is a preview of the 
> #this resource".
>
> Currently, only the contentType tells you something about what the 
> download is and that is admittedly not conveying this kind of detail. 
> It could, or another (new) column could... we should discuss a 
> detailed example. 



François 2019-10-11 wrote to Ada :
> This makes sense.
>
> There was already the project to create a new branch "associated_data" 
> in semantics
>
> Subranches would have been at least  the dataproduct types of Obscore 
> + maybe others
>
> associated_data:image, associated_data:TimeSeries
>
> we can have a further level in branching

Petr Skoda 2019-10-12 wrote to Ada :

> I am afraid that with trials to assign some semantics
> we are re-entering the problem of controlled vacabulary , its 
> management etc ...
>
>
> in fact the products that may be wanted by the community need not be just
> time  vs- some variable but also a (quite complex) function of time
>
> this is what we tried to address in our note 
> http://www.ivoa.net/documents/Notes/CubeDM/20190318/cubeDM.pdf
>
> e.g. fig.17 - the LIGO requirements - frequency vs strain
>
> or I can imagine in asteroseismology  people have power spectrum - i.e.
> Fourier transform vs frequency ....
>
> so the
>
> timeseries-X    has X from a kind of vocabulary ?
> In fact the visibility in radio is by definiton as well a complex 
> function of time ....
>
> could  we somehow use UCD (and add missing ones - e.g. the mirror 
> strain )?
>
> and their combination ? 

Petr Skoda 2019-10-12 wrote also to Ada :

>
> timeseries of images is in fact what we are presenting in our model in 
> Fig.~13 - to every point of a light curve is associated the link to 
> original image (here we were thinking about placing the provenenace )
>
> so the time series of images is just a normal time series (where 
> another "axis" paralel with e.g. flux is an axis of mappings (links to 
> images)
>
> As I have shown in Sydney 2015, the SPLAT , TOPCAT and ALADIN 
> understand this well and (Fig.15 in a note) displays the image. We can 
> even support cutouts ... 

François 2019-10-12 to Petr

>
> But the point raised by Ada was not to address the case where you want 
> to modelize this but more simply the case where you want to attach a 
> timeSeries dataset to a source in a catalog (GAIA use case)
> USING DAtaLInk
>
> Indeed, this is a vocabulary evolution problem. There is a proposal to 
> extend DataLink vocabulary (semantics field) see 
> (https://wiki.ivoa.net/twiki/bin/view/IVOA/UpdateDatalinkTerms)
>
>
> We should extend this  a little bit in the direction proposed by Ada
>
> Semantics working group will manage the procedure.
Ada 2019-10-12 to all
> Thanks for your comments and responses.
>
> Pat, content_type (the first field I looked into) tells you about the 
> format but not about the actual content. At least this is what I 
> understood.
> This is why I point to the ”semantics" field and adding a vocabulary 
> without having to add an extra column for that.
>
> Francois, Associateddata=timeseries-images,… something like that could 
> work.
>
> Petr, Francois is right in his comment as to what I refer to.
> I thought about the ucd usage of it, perhaps this is something to look 
> into.
>
> I guess the options/questions are:
> 1) add a column for that and add a vocabulary
> 2) use the field semantics and have a controlled
>
> In both cases it points to a vocabulary, and I would vote for 2 in 
> principle since it looks like it would require less effort.
>

Petr 2019-10-12 to all

> thanks for explanation - I was thinking about more complex cases than 
> he end point of datalink (e.g. animation of datacube)
> in this case we had a problem to describe the link of our example 
> pointing from individual photometry  point to the original image
> We used a SIAPLink with accesURL - as this was  understood by Aladin.
>
> the semantics makes sense and its controlled vocabulary as well..
> just I still see philosophical issue with the definition of time
>
> to fit powerspectra frequencies etc ....
>
> so it may be timeseries of (in principle) datacubes - spectra and 
> images are degenerated datacubes ...
>
> as an ordered set of entities - eg files in directory havin timestamp
>
>
> which is diffeent from the time transformation- like
> associate the timeseries with its power spectrum
>
> Kepler case e.g.
>
> how to work with this ?














Le 15/10/2019 à 17:28, François Bonnarel a écrit :
> Dear all,
>
>     Following the discussions last week and mainly the question of 
> tagging TimeSeries attached to sources in a catalog via DataLink I 
> make a proposal using the draft mechanism of "VEP" (vocabulary 
> enhancement proposal) proposed by Markus in the
>
> draft "Vocabularies in the VO version 2.0 -2019-07-15"
>
>    AssociatedDataSet is a new branch. Children are Associated-image, 
> Associated-spectrum, etc... as requested recently By Alberto Micol or 
> Carlos Rodrigo.
>
>   and Associated-TimeSeries. As requested by Ada during the interop I 
> added a few children to Associated-TimeSeries (so grand-children to 
> Associated-Data) : Associated-Lightcurve, 
> Associated-TilmeSeries-spectra, etc ...
>
> You will find this so-called   "VEP" here :
>
> https://wiki.ivoa.net/twiki/bin/view/IVOA/WebHome?topic=VEPs
>
> Cheers
>
> François
>
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ivoa.net/pipermail/voevent/attachments/20191015/e94d9934/attachment.html>


More information about the voevent mailing list