W3C home > Mailing lists > Public > public-egov-ig@w3.org > November 2010

Re: Geo in RDF (was Re: Censorship?)

From: Stuart Williams <skw@epimorphics.com>
Date: Tue, 16 Nov 2010 11:21:34 +0000
Message-ID: <4CE2693E.8090008@epimorphics.com>
To: bvillazon@fi.upm.es
CC: public-egov-ig@w3.org
Hello Boris,

On 15/11/2010 23:51, Boris Villazón Terrazas wrote:
> Hi Stuart, all
>
> On 15/11/2010 15:26, Stuart Williams wrote:
>> Hello Boris,
>>
>> On 13/11/2010 15:11, Boris Villazón Terrazas wrote:
>>> I only want to point out our ongoing work in the context of the 
>>> GeoLinkedData [1], in which  we are working with geospatial data. Basically, 
>>> we are using geo:lat/geo:long style of giving position in WGS84 coordinates. 
>>> A geospatial resource, for example a geoes:Provincia, has a geo:geometry, 
>>> and this geo:geometry consists of a set of geo:points , and each geo:point 
>>> consists of geo:lat and geo:long. You can find a figure describing this at [2].
>> I've got a minor question wrt to the use of geoes:order in the diagram at 
>> [2]. It seems to be giving the ordinal position of a Point in a LineString. 
>> It seems to me that there isn't a single value of geoes:order which applies 
>> to a given point... it depends on what LineString the point is being included 
>> in.
> We use the order for drawing the lines on the map, and we need to do it in the 
> right order. So, first, line from Point(x1,y1) to Point (x2,y2), then line 
> from Point(x2,y2) to Point(x3,y3) and so on.
> That's why we use the geoes:order
I think I understand the intention. My question really is how you use the same 
point in multiple curves - for example where two regions touch and share a 
common segment of their respective boundaries. For example reveals a number of 
points with more than one value for geoes:order:

    select distinct ?point ?order1 ?order2  where
    { ?point a <http://www.w3.org/2003/01/geo/wgs84_pos#Point>;
    <http://geo.linkeddata.es/ontology/orden> ?order1 ;
    <http://geo.linkeddata.es/ontology/orden> ?order2 .
       FILTER(?order1 < ?order2)
    } ORDER BY ?point ?order1 ?order2

e.g.:
         
http://geo.linkeddata.es/resource/wgs84/38.11124666521918_-2.519757266504941

Maybe that's intentional, its just that order within a linestring doesn't seem 
to me to be an invariant property of a point, more of its 'use' in defining a curve.

>>> A specific example, the resource Albacete Provincia at [3].
>>> Also, we have a browser at [4].
>>>
>>> I think this is related with the discussion.
>>> We can provide further information regarding the conversion from GML to RDF 
>>> we performed.
>>
>> I think that the details of your conversion are fairly clear in your 
>> presentation at http://www.slideshare.net/boricles/geo-upmv14boris
> Basically, the information from the Spanish Geographical Institute are stored 
> in Oracle 11 databases, and the geometry information in a blob column.
> We use the Oracle STO Util package [X]  for transforming the blob column value 
> into GML.
> Then, we use GeoTools [Y] and Jena [Z] for converting GML into RDF following 
> the model described in the diagram at [2]
>>
>> What I am looking for is a *widely* adopted practice, perferably 
>> backed/endorsed by a defacto or de-jure standards organisation. I don't think 
>> I have found such a beast except for the case of Point positions expressed as 
>> WGS 84 lat/long in which case there is widespread community practice in the 
>> use of the "Basic Geo (WGS84 lat/long) Vocabulary" [a].
>> [a] http://www.w3.org/2003/01/geo/
>>
> Our geometry points are compliant to geo: http://www.w3.org/2003/01/geo/wgs84_pos#
Yes, I can see that for points expressed in WGS 84. I happen to have some data 
that expresses points as ETRS89 lat/long(i.e. with a datum fixed to continental 
Europe). In truth for my application the difference between ETRS89 and WGS 84 
doesn't really matter - ~2.5cm per year continental drift. However, ideally I 
would like to express them as being ETRS89 which conveys (at least some level 
of) invariance over time within Europe. Yes, being pragmatic, I can express them 
as WGS 84 and maybe be as much as 25cm adrift by now - and in this particular 
case that's good enough for most/all practical purposes... but...
>> For curves (LineStrings), surfaces(Polygons) and  Boxes (Envelopes), the Geo 
>> OWL vocabulary at [b,c] which mirrors GeoRSS GML [d] seems to me to come 
>> close to being something that a common practice might develop around, however 
>> it does 'camp' in an opengis.org namespace without obvious endorsement from 
>> the OGC (unless I missed finding that, which is perfectly possible).
>>
>> [b] http://www.w3.org/2005/Incubator/geo/XGR-geo-20071023/#owl
>> [c] 
>> http://www.w3.org/2005/Incubator/geo/XGR-geo-20071023/W3C_XGR_Geo_files/geo_2007.owl
>> [d] http://www.georss.org/gml#GeoRSS_GML
>>
> We have checked them, but at that time we prefer to follow the gml:LineString, 
> however we can extend our model to include them.
It looks to me like you are using WKT encodings form [1]
>
>> I think Gannon was making a point about localized re-invention of vocabulary 
>> either being inevitable or maybe accidental - either way it mitigates against 
>> the widespread adoption of a common practice and potentially linits the reuse 
>> of data assets published using a particular local practice. It also mitigates 
>> against tooling, developing to support such things as spatial-indexes in 
>> SPARQL stores, because without the use of common vocabularies you loose the 
>> triggers that would induce index building - and responsive UIs seeking to 
>> display artefacts within spatial bounding boxes could well do with the help a 
>> spatial index could give.
> Why don't all of us  start to collect the experiences we have to publish and 
> consume geoespatial linked data?
Where I am at is that  I have data that I want to publish as RDF. Currently its 
only point data, but in general I'd expect to have line segments and polygons. 
The coordinates systems being used in the data are ETRS89 and UK National Grid. 
Yes... I can convert to WGS 84 (in some cases that is a time dependent 
conversion) - but I'd like to respect the data's original form.

What I don't see is firm guidance on how to publish anything other than WGS84 
point positions as RDF. Yes... I could follow one of a number of [previously 
mentioned] emerging patterns and 'it-would-be-nice' (tm) for there to be some 
convergence of practice. I could make up my own scheme, but that would kind of 
miss the point and only add to the problem.

> Thoughts? suggestions?
I guess my ideal would be for guidance to emerge from OGC, ISO TC 211 and/or 
W3C.  That may be happening... but I don't have a good handle on what activity 
(if any) is focussed on this topic.
> Best
>
> Boris
>
BR

Stuart
--
>
> [X] SELECTTO_CHAR(SDO_UTIL.TO_GML311GEOMETRY(geometryColumn))
>
> AS Gml311Geometry
>
>       FROM "DB"."TABLE" c
>
>       WHEREc.Name='Arroyo'
>
> [Y] http://www.geotools.org/
> [Z] http://jena.sourceforge.net/
>>>
>>>
>>> geo: http://www.w3.org/2003/01/geo/wgs84_pos#
>>> [1]: http://geo.linkeddata.es/
>>> [2]: http://mccarthy.dia.fi.upm.es/challenge/example1.png
>>> [3]: http://geo.linkeddata.es/page/resource/Provincia/Albacete
>>> [4]: http://geo.linkeddata.es/browser/
>>>
>>
>> BR
>>
>> Stuart
> --
> Boris Villazón-Terrazas
> Ontology Engineering Group (OEG)
> http://www.oeg-upm.net
>
> Departamento de Inteligencia Artificial
> Facultad de Informática
> Universidad Politécnica de Madrid
> Campus de Montegancedo, s/n
> Boadilla del Monte - 28660 Madrid, Spain 


-- 
Epimorphics Ltd                        www.epimorphics.com
Court Lodge, 105 High Street, Portishead, Bristol BS20 6PT
Tel: 01275 399069

Epimorphics Ltd. is a limited company registered in England (number 7016688)
Registered address: Court Lodge, 105 High Street, Portishead, Bristol BS20 6PT, UK
Received on Tuesday, 16 November 2010 11:22:18 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 16 November 2010 11:22:19 GMT