[seek-kr] RE: Please tear this apart...

[Matt Jones]

Rich and Shawn,

This has been a great thread -- thanks for pursuing it so actively.  I'd 
like to note that a lot of what you are doing is taking the current work 
that is implicit in eml-attribute and modeling it as an ontology rather 
than an XML schema.  I think it would be important to NOT duplicate the 
information that is already in EML (e.g., precision, domain, unit) for 
every attribute, but rather find ways to leverage this previous work 
(e.g., by importing EML in a way that a logic engine can treat it as if 
it were an ontology).

Precision is a good case in point: there is an active dicussion on 
precision on eml-dev, and they are pointing out some excellent issues 
about how precision is measured in the real world, and how data is 
represented in the real world.   Our current model of precision in EML 
clearly does not model how scientists regularly think of precision and 
how they model their data may in fact prevent effective labeling without 
row-level addressing for data sources.  Up until now, EML has tried to 
apply all metadata to the schema elements themselves (e.g., entities and 
attributes).  Its becoming clear that we will need to be able to 
reference observations directly to attach metadata to specific 
observations within an attribute (e.g., in one data set precision in a 
relative humidity varies with temperature in a non-continuous way, and 
so there is no single precision value for the relative humidity 
attribute).  Check out the eml-dev archives for this precision thread 
that's been ongoing.

Matt

Rich Williams wrote:
> Replies are interspersed below
> 
> Rich
> 
> 
>>-----Original Message-----
>>From: seek-kr-admin at ecoinformatics.org
>>[mailto:seek-kr-admin at ecoinformatics.org]On Behalf Of Shawn Bowers
>>Sent: Monday, September 08, 2003 11:46 AM
>>To: seek-kr at ecoinformatics.org
>>Subject: [seek-kr] RE: Please tear this apart...
>>
>>
>>On Mon, 8 Sep 2003, Rich Williams wrote:
>>
>>
>>>Of course the units ontology I wrote is only a small piece of
>>
>>what you are
>>
>>>asking about, but none the less, I think it is an ontology by
>>
>>most common
>>
>>>computer science definitions and is quite similar to the units
>>
>>ontology that
>>
>>>you posted.  (Of course this begs the question of what is the difference
>>>between an ontology and a database schema - in general, my
>>
>>answer is 'not
>>
>>>much'...  See for example p156 in Sowa's Knowledge Representation book)
>>>
>>>I've checked in a slightly revised version of it, and also a
>>
>>new figure that
>>
>>>includes the slot relationships that hopefully makes its
>>
>>structure clearer
>>
>>>(I omitted them in the original for clarity).  I also isolated
>>
>>my attempt at
>>
>>>describing measurements from its extension to ecological measurements in
>>>particular and called the ontology MeasurementBase.  It's in the same
>>>seek/projects/kr/ontologies folder as the Units ontology.  It
>>
>>includes the
>>
>>>units ontology, and so gives a picture of how units relate to
>>
>>measurements
>>
>>>etc.  There's a figure, MeasurementBase.gif, of the ontology.
>>
>>I actually downloaded the ontology into Protege, and used
>>Graphviz to view
>>it :)
>>
>>I think it looks great.  I only have a few suggestions/questions...
>>
>>It seems like the classes that end in "Concepts" are really meta classes
>>(and could probably go away). For example, Measurement and Measurement
>>Context are particular Measurement Concepts. In other words, if I have an
>>instance of a Measurement (e.g., 5 g/cm), it probably shouldn't be an
>>instance of a Measurement Concept too (which is assumed by the IS-A).
>>
> 
> 
> The classes that end in "Concepts" could indeed go away.  I created them as
> an organizational convenience.  They basically top level containers designed
> to make browsing the ontologies easier, an issue especially when the
> ontology gets large.
> 
> 
>>Is ReferenceValue something specific to your Food Web
>>application? I don't
>>really understand what a referenced value signifies.
>>
> 
> 
> It's not used yet, I put it in as a place holder that can get developed in
> the future.  It's meant to be a pointer of some kind to a data value, to be
> used for example when data values are stored in a separate file.  So if the
> data was in an XML file it could be a file name and an xpath location path.
> There's a lot of work to do here - for example, data are typically in data
> sets that consist of one or more files, and there's no need to redundantly
> store file names in every data reference.
> 
> 
>>It looks like you included RatioMeasurement as an IntervalMeasurement. It
>>seems like it should have a property that further defines its units... or
>>that it is created from two measurements?  How would Proportions and
>>Densities/Concentrations be represented?
>>
> 
> 
> Your confusion is due to some confusing terminology.  RatioMeasurement is
> 'ratio' in the sense of nominal-ordinal-interval-ratio used by statisticians
> and described I believe in the eml-attribute documentation.  It does not
> mean that the measurement is a ratio of two quantities (though it could be).
> Ratio in that sense, as well as densities, concentrations etc would all be
> represented using the Dimension, which is defined in the Unit (note that
> IntervalMeasurement has a Unit property that is inherited by the
> RatioMeasurement class).
> 
> 
>>It seems like Precision and Accuracy need to be related to a Measurement?
>>
> 
> 
> I think that the correct relationship is that precision and accuracy are
> properties of NumericValues.  It's not in there yet though...
> 
> 
>>Shawn
>>
>>
>>
>>>Rich
>>>
>>
>>_______________________________________________
>>seek-kr mailing list
>>seek-kr at ecoinformatics.org
>>http://www.ecoinformatics.org/mailman/listinfo/seek-kr
> 
> 
> _______________________________________________
> seek-kr mailing list
> seek-kr at ecoinformatics.org
> http://www.ecoinformatics.org/mailman/listinfo/seek-kr

-- 
-------------------------------------------------------------------
Matt Jones                                     jones at nceas.ucsb.edu
http://www.nceas.ucsb.edu/    Fax: 425-920-2439    Ph: 907-789-0496
National Center for Ecological Analysis and Synthesis (NCEAS)
University of California Santa Barbara
Interested in ecological informatics? http://www.ecoinformatics.org
-------------------------------------------------------------------