In order to release the Social Semantic Web and solve dataportability issues, there is a need from Web 2.0 providers to open their content and deliver it in a machine-readable way. Modeling it with vocabularies such as FOAF, SIOC, as well as reusing data from the Linked Data initiative like DBpedia or GeoNames can help to achieve this task. The goal of this demo is to showcase an RDF exporter for Flickr pro les that acts such a way.
Describing people, user accounts and content using
Semantic Web technologies o ers standardized ways to query,
browse and reuse it. Moreover, it lets this data enter the
Linked Data Web rather than being part of a closed-world,
proprietary, data silo. The Flickr wrapper1 described here
provides a way to model a Flickr user account in RDF using
well-known vocabularies like FOAF[
The exporter mainly uses the following vocabularies and reference datasets to export user accounts:
FOAF - as a standard vocabulary to model people and their relationships, FOAF is used to represent the owner of a Flickr account (foaf:Person), his proper
ties and links with other people (foaf:knows) retrieved
from his Flickr contacts;
SIOC - as a model to describe activities of online
communities, recently accepted as a W3C submission, SIOC
is used to represent user accounts (sioc:User) related
to their owner (foaf:holdsAccount) and groups they
may belong to (sioc:UserGroup and sioc:member_of).
Moreover, the SIOC types module4 is used to represent
user galleries (sioct:ImageGallery and sioc:owner_of);
GeoNames5 - since Flickr allows any user to mention
his location, the script relies on the GeoNames
webservice6 to nd the related place URI, and thus uses
the geonames:locatedIn property and URI of
existing geonames:Feature resource to link a user to its
location. Since most geonames:Feature instances are
linked to DBPedia[
Each exported people, user, and group is assigned a
speci c non-informative resource URI. Each URI is
deferencable, and uses content-negociation principles to redirect
either to the Flickr.com user page if browsed with an HTML
browser, or to the RDF le that contains all the statements
about the user, account or group if browsed with an RDF
browser. Those redirection use 303 and 200 HTTP codes
to be compliant with Semantic Web URI principles [
Reusing Linked Data principles[
URI created with the exporter is to combine the use of owl:sameAs and rdfs:seeAlso. Thus, a person that already has a FOAF URI can identify himsemlf as the same than the one described by another URI in a Flickr pro le. It also allows tools supporting OWL inference to make such identi cation when querying various sets of RDF data.
Moreover, in case the user do not have any existing FOAF pro le, this script gives him a URI (and a pro le) that he can the reuse in other Semantic Web applications like Revyu.com8 to be uniquely identi ed in the Semantic Web.
Created data can be used as any RDF data, for exemple browsed with the Tabulator9 extension for Firefox, as shown on Fig.1. Moreover, people can use SPARQL to run query over the data, either on the y or by crawling it, and in case they already have RDF data exported from services like Twitter10 or Facebook11, may merge them and query or view their complete open and machine-readable social network in a uniform manner. Fig.2 displays the output of a simple script12 that queries a FOAF pro le to retrieve all owl:sameAs URIs and related RDF documents for a given user to nd all his relationships from di erent social networks. It provides a simple but e cient use-case to see how Semantic Web technologies can achieve the goals of social network and data portability.
In order to respect the Flickr API terms of use and user privacy, only publicy available data can be exported. Yet, in case a user switch his pro le from private to public, he currently must recreate his exported pro le (in case he wants the old one to be removed), since there is not direct way for the script to be informed of Flickr users changes at the moment. 7.
Di erent wrappers to create RDF data from existing sources have been mentioned in this paper. Other exporters may include the AudioScrobller wrapper13 or OpenLink Virtuoso sponger14, a more generic way to extract RDF data from non-RDF sources. However, there is not yet agreement on how to export such data and to combine vocabularies. This demo may be used as a basis for best practices to do so.
Finally, regarding Flickr, the Flickcurl API15 provides a way, among various operations, to get RDF triples from any Flickr picture. A script to interlink this data and the currently described exporter may be available in the future. 8.
The work described here is personal work, and is not afliated with Flickr.com in any way. 13http://dbtune.org/last-fm/ 14http://www.openlinksw.com/virtuoso/Whitepapers/html/ VirtSpongerWhitePaper.html 15http://librdf.org/ ickcurl/