W3C home > Mailing lists > Public > semantic-web@w3.org > August 2008

Re: New Exif/RDF vocabulary

From: Tim Berners-Lee <timbl@w3.org>
Date: Fri, 8 Aug 2008 14:07:41 -0400
Cc: SW-forum Web <semantic-web@w3.org>, Ivan Herman <ivan@w3.org>, Norman Walsh <ndw@nwalsh.com>
Message-Id: <4025FD0F-F10F-475B-AE9E-BEB35FF08D63@w3.org>
To: KANZAKI Masahide <mkanzaki@gmail.com>

Kanzaki,

I have only now looked at it.  The tools look nice.

I have one comment about the modeling.

Looking at for example

  http://www.kanzaki.com/test/exif2rdf?u=http://web.canon.jp/Imaging/ixy30/SPL/IMG_0138.JPG

there is what seems to be an extra arc "exifdata" to a bnode labelled  
"Exif Image File Data", and then the exposureTime is a property of the  
exifdata, not of the original phototo.
The arc seems to be used to group the data in the file, but I would  
find it more useful I think
to define the exposureTime as a property of the picture itself.

I like the modeling of  the various enumeration types I think make  
sense, such as YCbCr vs sRGB etc.

I would be more inclined to model units using predicates. Instead of  
(in N3)

             exif:focalPlaneResolutionUnit
                exif:inches;
             exif:focalPlaneXResolution
                "9846.15384615";
             exif:focalPlaneYResolution
                "9846.15384615";

I would be inclined to use

             exif:focalPlaneXResolution
                [ exif:inches "9846.15384615" ];
             exif:focalPlaneYResolution
                [ exif:inches "9846.15384615"];

You can the for example deduce what the exif:focalPlaneYResolution of  
something is in meters without any contradictions.


             exif:focalPlaneXResolution
                [ exif:inches "9846.15384615" ; unit:m "250.0"];
             exif:focalPlaneYResolution
                [ exif:inches "9846.15384615"; unit:m "250.0"];

(Hmm .. in this example it looks as though the meters was probably the  
original figure!)

or even

             exif:focalPlaneXResolution
                [ exif:inches 9846.15384615 ; unit:m 250.0];
             exif:focalPlaneYResolution
                [ exif:inches 9846.15384615 ; unit:m 250.0];


I made a simple suitable units ontology by the way at http://www.w3.org/2007/ont/unit

Should a group be standardizing the exif ontology, do people think?
How stable is the http://www.kanzaki.com/ns/exif namespace?
Would you like a w3.org namespace for it?

Keep up the good work!

Tim
______________________________
excerpt from the exit example mentioned above:

  <foaf:Image rdf:about="http://web.canon.jp/Imaging/ixy30/SPL/IMG_0138.JPG 
">
   <dc:date>2003-01-18T16:07:30</dc:date>

<!-- Metadata from Exif section -->
   <make>Canon</make>
   <model>Canon IXY DIGITAL 30</model>

   <orientation rdf:resource="http://www.kanzaki.com/ns/exif#right- 
top"/>
   <xResolution rdf:datatype="&xsd;decimal">180</xResolution>
   <yResolution rdf:datatype="&xsd;decimal">180</yResolution>
   <resolutionUnit rdf:resource="http://www.kanzaki.com/ns/ 
exif#inches"/>
   <dateTime>2003-01-18T16:07:30</dateTime>
   <yCbCrPositioning rdf:resource="http://www.kanzaki.com/ns/exif#centered 
"/>
   <exif-info>

    <Exif_IFD rdfs:label='Exif Image File Data'>
     <exposureTime rdf:datatype="&xsd;decimal">0.0025</exposureTime>
     <fNumber rdf:datatype="&xsd;decimal">3.5</fNumber>
     <exifVersion>0220</exifVersion>



[1] http://www.w3.org/2001/sw/BestPractices/MM/interop.html

On 2007-12 -13, at 11:50, KANZAKI Masahide wrote:

>
> Hello all,
>
> Several years after my release of Exif RDF vocabulary [1], I updated
> the schema with some significant changes.
>
> The old vocabulary was defined as RDF schema and no rdfs:domain and
> rdfs:range were provided for property descriptions. While many of Exif
> tag values come just as numbers, how to describe those values in RDF
> was not cleary defined.
>
> This time, I redefined the vocabulary as OWL ontology, so that tags
> whose values have pre-defined interpretation are defined as
> owl:ObjectProperty, and their possible values as resources denoted by
> URIs. Also, domain of properties are defined (via super properties) so
> as Exif's IFD structure to be retained in the RDF data.
>
> Although those changes made the new definition incompatible with the
> old one, I would keep its namespace URI [1 + '#'] intact. If you need
> old definition vocabulary (which might be easier to use in
> hand-writing), its copy is available at W3C with somewhat authentic
> URI [2]. I will finalize this change within a week unless any
> objection (e.g. assign different URI for new definition).
>
> For users convenience, the new definition comes with a JSON file [3]
> which provides mapping from Exif tags/values to RDF
> properties/value-resources. A sample Web application to extract RDF
> from image file [4] now uses the new definition.
>
> Comments, suggestions are welcome, of course.
>
> thanks,
>
> [1] http://www.kanzaki.com/ns/exif
> [2] http://www.w3.org/2003/12/exif/ns
> [3] http://www.kanzaki.com/ns/exif-tags.json
> [4] http://www.kanzaki.com/test/exif2rdf
>
>
> -- 
> @prefix : <http://www.kanzaki.com/ns/sig#> . <> :from [:name
> "KANZAKI Masahide", "神崎正英"; :email "mkanzaki@gmail.com"].
>
Received on Friday, 8 August 2008 18:17:22 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 21:45:24 GMT