[Seavox] Food for thought -- Problem 2

[Roy Lowry]

Hi John,

Yes, my frustration was pushing me to suggest something close to ISO sacrilege.  However, it's not as bad as you make out because it's the structures within ISO19115 to incorporate non-ISO keyword lists (the MD_Keywords class) that I'm talking about and not the ISO code lists incoporated inside the standard.

The reason for the frustration is that whilst the full ISO19115 CI_Citation class (not the cut-down version in the Marine Community Profile) gives me everything I need to store all the bits of a URI to a versioned list (either to Luis's specification or the cut-down specification I'm currently using until I put the infrastructure together for publication of list content governance), there is nowhere to put the extra bit of information needed (the term key) to turn a list URI into a term-within-list URI.

To move to your example.  Basically you have two technical governances (BODC/MMI) serving a list under external content governance (CF).  Which is authoritative?  The answer is that this totally depends on the view of the content governance authority.  If they are satisfied that both BODC and MMI serve correct, maintained versions then both are authoritative (and should be identical). The ISO CI_Citation class allows the content governance authority and technical governance to be identified thereby providing the basis for verification.

Cheers, Roy.  

>>> John Graybeal <graybeal at mbari.org> 2/26/2007 5:55 pm >>>
Roy,

Re Problem 2 ---

I think the 'problem' here in ISO may be the lack of a specific process for mapping the keyword term/citation to a unique reference.  To be clear I need to make up an artificial example: Let's say I claim the thesaurus is CF vocabulary terms, but I don't specify a URI for that reference.  If two organizations (BODC and MMI) both serve CF vocabulary terms, which one is authoritative from the ISO usage standpoint?  In the ISO case they may have specified the citation provisions in more detail (sorry I haven't had a chance to review that yet -- can someone provide an authoritative answer?).

You can solve this problem definitively for your particular case in two ways: 1) by citing a citation class using a URI that you know will resolve the keywords to a specific entry, or 2) by citing a complete URI as your keyword (in which case the citation class becomes irrelevant, and everyone else freaks out about your non-ISO keywords.  I mention the latter case only because your last paragraph seems to say this -- may be a typo.

Option 1 seems an elegant solution, and may help set the standard for how everyone does it (and I'd encourage it be added to the marine profile Greg Reed put together).  Maybe someday ISO will be in a position to take the next step and be a vocabulary serving authority, but in the meantime I think you, MMI, and others will have to step up to fill that gap.  Depending on your goals you could include the specific date of the vocabulary in the citation, or just declare 'latest' as the date -- your approach (the former) seems best.

May I suggest you (or others -- you probably know of these) take a look at Luis Bermudez' recommendations  on how to create URI references to OWL vocabularies (http://marinemetadata.org/owluriguidelines) -- it is fairly mature but still needs review and improvement.  To the extent you find these practices worth following/endorsing (or correcting), maybe we can get a groundswell of enthusiasm going.

john


At 4:30 PM +0000 2/26/07, Roy Lowry wrote:
>Dear All,
>
>Just thought I'd share a couple of my current quandries with you to see if any pearls of wisdom out there can ease my suffering!
>
>Problem 1
>
>Consider the AMT dataset - repeated measurements taken along 20W (i.e. straight down the middle of the Atlantic Ocean).  The CTD temperature data in this are marked up with the BODC Parameter Usage Vocabulary term 'Temperature of the water column by CTD'.  The plan is to use this markup to automatically generate metadata records.  So, we first of all convert the usage term to a discovery term, resulting in 'Temperature of the water column'.  Now for ISO19115 we need Topic Categories and because the discovery term covers datasets from both fresh and salt water (to get around the problem of datasets running up estuaries and rivers) we end up with a mapping relationship to both 'oceans' and 'inlandWaters'.  So, we end up associating our mid-Atlantic dataset with the ISO 'inlandWaters' topic, which to me is totally daft.  However, the logic used to derive the relationship (a SKOS minorMatch mapping) seems sound. 
>
>I guess this is exactly the same problem of false indication in the parameter dimension that is encountered in the spatial dimension when using rectangular boxes to describe dataset spatial coverage.  However, to me the spatial case seems acceptable whereas the parameter case does not.  Does this mean that problem is that my perception of 'acceptable' is based on common sense and not logic?
>
>Problem 2
>
>The ISO19115 keyword class includes a keyword type label, a list of actual keyword terms and a citation class to document the source thesaurus. There is no specific location for keyword URIs, although these could be built from the citation class and the keyword term providing a representation of the thesaurus version used containing both terms and keys was available.  Is this adequate or should we be pressing for inclusion of an explicit keyword URI storage location in ISO19115/19139 profiles?  I raised this issue expressed in terms of 'keys' rather than 'URIs' for the ISO19115 Ocean Profile with Greg Reed back in September, am coming to realise that when expressed in terms of URIs the case is much stronger.
>
>I guess another possibility would be to use a URI that unambiguously described the keyword text term (such as  http://vocab.ndg.nerc.ac.uk/term/P021/12/TEMP/term/ for the keyword text from entry (key+term+abbreviation+definition) http://vocab.ndg.nerc.ac.uk/term/P021/12/TEMP/ from thesaurus http://vocab.ndg.nerc.ac.uk/term/P021/12/) as the ISO19115 keyword element content. Note 'P021/12' specifies version 12 of controlled vocabulary list P021. Or is being too creative?
>
>Cheers, Roy.
>
>
>_______________________________________________
>Seavox mailing list
>Seavox at mailman.nerc-liv.ac.uk 
>http://mailman.nerc-liv.ac.uk/mailman/listinfo/seavox 


-- 
----------
John Graybeal   <mailto:graybeal at mbari.org>  -- 831-775-1956
Monterey Bay Aquarium Research Institute
Marine Metadata Initiative: http://marinemetadata.org   ||  Shore Side Data System: http://www.mbari.org/ssds