=Paper=
{{Paper
|id=Vol-368/paper-5
|storemode=property
|title=Scripting User Contributed Interlinking
|pdfUrl=https://ceur-ws.org/Vol-368/paper6.pdf
|volume=Vol-368
}}
==Scripting User Contributed Interlinking==
https://ceur-ws.org/Vol-368/paper6.pdf
Scripting User Contributed Interlinking
Michael Hausenblas1 , Wolfgang Halb1 , and Yves Raimond2
1
Institute of Information Systems and Information Management,
JOANNEUM RESEARCH, Steyrergasse 17, 8010 Graz, Austria
firstname.lastname@joanneum.at
2
Centre for Digital Music,
Queen Mary, University of London, UK
yves.raimond@elec.qmul.ac.uk
Abstract. When building a linked-data dataset for humans and ma-
chines, a range of issues emerges. In this paper we discuss our findings
regarding the implementation of riese (http://riese.joanneum.at), the
RDFized and interlinked version of the Eurostat data. The contribution
of our work is twofold: On the one hand we propose a new way of cre-
ating semantic links, labelled as User Contributed Interlinking, on the
other hand we discuss integration issues regarding Ajax and embedded
RDF-metadata.
1 Motivation
In early 2007 the Linking Open Data (LOD) community project has been launched
within the W3C Semantic Web Education and Outreach (SWEO) group. The
LOD community project bootstraps the Semantic Web by publishing datasets
in RDF [1] on the Web. By creating large numbers of typed links between
datasets [2, 3] Semantic Web application development is fostered.
Several issues emerge when building an LOD dataset; from the schema level—
that is how to map from, e.g., a relational schema to an RDF Schema—to the
proper and meaningful assignments of URIs to entities. One key success factor is
the used interlinking method. Several approaches exist for semantically linking
data.
With riese (“RDFizing and Interlinking the EuroStat Data Set Effort”) [4]
we have contributed to the LOD cloud by adding the Eurostat data. The im-
plementation of riese heavily depends on scripting languages such as PHP and
JavaScript. In this paper we report on our findings when implementing the riese
dataset. We introduced a new way of enriching datasets called “User Contributed
Interlinking” (UCI), which is a Wiki-style approach enabling users to add seman-
tic (that is: typed) links between data items on a URI-basis.
The paper is structured as follows: In section 2 we briefly introduce the LOD
principles and discuss the current state of the LOD datasets. Then, in section 3
we explain the riese implementation, including its UCI-interface and Web 2.0
issues. In 4 we discuss the generalisation of the UCI. Finally, we conclude our
findings in section 5.
2 A Linking Open Data Overview
According to Tim Berners-Lee3 it is desirable to interlink datasets, thus allowing
the discovery of more data. Interlinking is one of the key factors that made the
(hypertext) Web so successful. The same principle holds for the Semantic Web;
rather than creating (untyped) hyperlinks between different documents, semantic
links are used to interlink data items from different sources.
ECS Sem-
South- Web-
ampton Central
updated
Music- Doap-
Audio- space Flickr
brainz
Scrobbler QDOS exporter SIOC
profiles
BBC BBC Magna-
Onto- SW
Later + John tune
Jamendo world
Conference
TOTP Peel FOAF Corpus
profiles Open-
Guides
Geo-
names Revyu
DBpedia RDF Book
US
Census Mashup
World
Data NEW! Fact- DBLP
book lingvoj
riese Berlin
NEW!
RKB
Euro- Explorer
stat
flickr
Gov- Wiki- Open wrappr
Track company Cyc
DBLP
Hannover
W3C Project
WordNet Guten-
berg
Fig. 1. The Linking Open Data dataset cloud in early 2008.
The Linking Open Data (LOD) project [2] is an open, collaborative effort
carried out in the realm of the W3C SWEO4 community projects initiative.
Notable landmarks within the LOD community project include the publication
of common, real-world datasets such as Wikipedia, Geonames and Musicbrainz.
Currently, the project includes over 30 different datasets (cf. Fig. 1, by courtesy
of Richard Cyganiak5 ), ranging from rather centralised ones (such as DBpedia)
to those that are very distributed (cf. the FOAF-o-sphere).
From one billion triples and 250k links in mid-2007 the LOD dataset has
grown to over two billion triples and 3 million links in early 2008, representing
a steadily growing, open implementation of the Linked Data principles detailed
below. It should be noted that we do not view this dataset as a fixed, delineated
3
http://www.w3.org/DesignIssues/LinkedData.html
4
http://www.w3.org/2001/sw/sweo/
5
http://richard.cyganiak.de/2007/10/lod/
or closed entity, but rather a snapshot of a major data ecosystem within the
Semantic Web at this point in time.
Linked Data Principles The linked data principles read as follows:
1. All items should be identified using URI references (URIrefs)6 , which implies
that ideally no blank nodes are used7 ;
2. All URIrefs should be dereferenceable—using HTTP URIs allows looking up
the items identified through URIrefs; see also the so called “http-range-14
TAG finding”8 );
3. When looking up an URIref—that is, a property is interpreted as a hyperlink—
it leads to more data, which is usually referred to as the follow-your-nose
principle [5];
4. Links to other URIrefs should be included in order to enable the discovery
of more data.
Interlinking [6] describes how to publish linked data, and further discusses the
two basic approaches for creating links to other datasets. Generally speaking,
the RDF links can either be set manually or generated by automated linking
algorithms for large datasets. For the latter case Raimond et.al. [7] have shown
that simple interlinking algorithms produce rather poor results.
Naive approaches trying to perform a simple literal lookup are likely to fail;
for instance, when trying to interlink data from the geographical domain with
Geonames it is possible to do a simple literal lookup using the search facility
provided by Geonames. However, when querying for the city Vienna almost 20
results will be returned as there exist that many cities named Vienna around the
world. Advanced approaches such as described in [7] are needed to disambiguate
similar matches and finally create appropriate interlinks. Still, there is no guar-
antee that the automatically generated interlinks are truly relevant. Moreover
the automated process is also restricted to predefined datasets implying that
only a subset of the data available on the Semantic Web is considered when
looking for potential interlinks.
3 riese: Scripting LOD for humans and machines
With riese (launched in early 2008), we aim at offering a Semantic Web version
of the publicly accessible data provided by the Eurostat data source, for both
humans and machines. We currently serve some 3.6 million RDF triple in total,
including the interlinking to Geonames data. There is ongoing work to cover
the total Eurostat data (yielding some 4 billion RDF triple) and extending the
interlinks to DBpedia, WordNet and other LOD data sets.
In Fig. 2 the riese system architecture is depicted. The data from Eurostat is
converted into RDF/XML using SWI-Prolog, and dumped into the file system.
6
http://www.w3.org/TR/rdf-concepts/#section-Graph-URIref
7
http://iandavis.com/blog/2007/03/bnodes-out
8
http://www.w3.org/2001/tag/doc/httpRange-14/HttpRange-14.html
This is to say that for each table—in tab separated values (TSV) format—
from the Eurostat download9 a corresponding RDF/XML (content.rdf) file,
holding the statistical data, exists. The riese core schema is modelled using RDF-
Schema [8] and comprises three main classes: riese:Dataset, riese:Item and
riese:Dimension. A dataset is the logical container of either more sub-datasets
(related via skos:narrower) or data items. We refer to [4] for further details on
the modelling issues of the schema.
An Apache 2 Server along with a set of PHP scripts is used to render the
pages in XHTML+RDFa [9]. The riese front-end is very light-weight; some 450
LOC in PHP and ca. 130 LOC in JavaScript were necessary to create a pleasant
yet functional Web-based user interface.
Fig. 2. The system architecture of riese.
Additional to the statistical data available for both humans and machines,
riese offers another novel feature: we have implemented a User Contributed In-
terlinking (UCI) interface, discussed in the following.
9
http://europa.eu/estatref/download/everybody/
3.1 UCI in riese
The User Contributed Interlinking (UCI) part of riese, the UCI-interface, can
be understood as an agent in the sense of [10]. The UCI-interface allows to list,
add, and remove user-contributed semantic links from each of the statistical data
items10 .
Operation Query String
list semantic links of the ?src=sURI
data item sURI
add a semantic link to the ?src=sURI&property=pURI&target=tURI
data item sURI
remove a semantic link from ?src=sURI&property=pURI&target=tURI&remove
the data item sURI
Table 1. Supported operations of the UCI-interface.
The operations supported by the current version of the UCI-interface are
listed in Table 1. Note that the base service URI http://riese.joanneum.
at/interlinking/uci-interface.php is assumed. With an additional format
parameter the output format can be controlled. The default format is XHTML,
an RDF/XML representation can be obtained using format=RDF.
To avoid concurrent editing a simple lock mechanism has been implemented.
In case two users simultaneously want to add a semantic link to a data item, an
according “please-hold-the-line” message is displayed.
It has to be noted that the UCI data is kept in a separate document—that
is, a separate RDF/XML document, uci-store.rdf, per data item—in order
to allow updates independently from statistical-data updates.
1 @PREFIX rdfs : < http :// www . w3 . org / 2 0 0 0 / 0 1 / rdf - schema # > .
2 @PREFIX foaf : < http :// xmlns . com / foaf /0.1/ > .
3 @PREFIX riesed : < http :// riese . j o a n n e u m . at / data / >.
4
5 < riesed : economy >
6 rdfs : seeAlso < http :// www . unece . org / Welcome . html > ;
7 foaf : topic < http :// dbpedia . org / r e s o u r c e / Economy > .
Listing 1.1. An example result from a UCI query.
To obtain, for example, an RDF representation of the UCI data for the
data item http://riese.joanneum.at/data/economy, one would use the query
10
http://riese.joanneum.at/data/
string ?src=http://riese.joanneum.at/data/economy/&format=RDF. The re-
sult (rendered in RDF/N3 for better readability, here) would then be as shown
in listing 1.1.
The HTTP-GET-interface of the UCI itself contains some 270 lines of code
(LOC) in PHP, extensively making use of the RAP library11 .
UCI User Interface On the client side we have implemented an user interface
that controls the UCI-interface using Ajax (the UCI-UI). The Yahoo! User In-
terface Library (YUI)12 has been utilised for panels, events, etc. but also for
the asynchronous communication. The UCI data is merged into the UCI user
interface at rendering time. Fig. 3 shows the UCI user interface “launch pad”:
For each data item a user may choose to add semantic links using the “I know
more” button, effectively launching the UCI-UI.
Fig. 3. UCI in riese - I.
In Fig. 4 the main UCI panel is depicted. Users can view, add, and remove
semantic links with it. Note how the subject of the RDF statement is implicitly
set to the data item from which it has been fired. Currently three semantic link
types (properties) are supported (owl:sameAs, rds:seeAlso, and foaf:topic).
We decided to control this part of the RDF statement as well strictly to (i) make
it easier to use for the average user from the street, and (ii) to avoid issues when
following-your-nose. Finally, the object of the RDF statement is the open part
of the UCI data. With open we mean that is is up to the user to determine what
URI to paste in. However, people are encouraged to use URIs pointing to RDF
11
http://www4.wiwiss.fu-berlin.de/bizer/rdfapi/
12
http://developer.yahoo.com/yui/
Fig. 4. UCI in riese - II.
(or GRDDL-able) resources. The UCI implementation is still in an early stage;
further investigations both regarding scalability and usability are under way.
3.2 Issues in Web (2.0) Environments
Issues with embedded metadata and Ajax As described in [11] issues such as
the do-not-repeat-yourself principle, the locality of structured data, or self-
containment of descriptions need to be addressed when embedding metadata
in (X)HTML. As these are generic issues, they are not limited to a specific
methodology or technology, such as microformats13 , eRDF14 , or RDFa. For a
deeper discussion of these issues the reader is referred to [12].
We have encountered issues with the in-place creation of RDFa in the utilised
Ajax framework (YUI). Whenever rendering the metadata—expressed in RDFa,
in our case—directly in the DOM, a non-DOM-based extractor is not aware of the
RDF, hence unable to make use out of it. Take for example the RDFaDistiller15 ,
a REST-based, conforming RDFa-processor. When RDFaDistiller fetches the
content from a data item it is not able to access the DOM-only parts, hence they
are lost. This seems to be a general problem when using embedded metadata
along with dynamic content. We are not aware of a fix allowing a generic solution.
We note, however, that for example in a Last Call comment to RDFa this is
has been recorded as a known issue to be addressed in future versions of this
standard16 .
Access of Semantic Web data sources Another integration issue turned out to be
the access of RDF-based resources. The use case in riese reads as follows: When
new data is available, one way to signal this is to subscribe to a news feed. We
13
http://microformats.org/
14
http://research.talis.com/2005/erdf/wiki/Main/RdfInHtml
15
http://www.w3.org/2007/08/pyRdfa/
16
http://www.w3.org/2006/07/SWD/track/issues/114
chose Atom [13] as the news feed format, as an corresponding RDF vocabulary
(AtomOwl [14]) exists.
1 < body about = " http :// riese . joanneum . at / updates " instanceof = " awol : Feed " >
2 < div rel = " awol : title " instanceof = " awol : Content " >
3 < span property = " awol : body " > updates
4
5 < div id = " main - updates " >
6 < ul rel = " awol : entry " instanceof = " awol : Entry " >
7 < li rel = " awol : title " instanceof = " awol : Content " >
8 < span property = " awol : body " > Compensation of employees - NACE J - K -
Current prices - Millions of euro - SA :
9 < span rel = " awol : link " instanceof = " awol : Link " >
10
11 http :// riese . joanneum . at / data / na075
12
13
14
Listing 1.2. An AtomOwl data update example in XHTML+RDFa.
On the riese updates page (http://riese.joanneum.at/updates/) the data
news feed is made available in AtomOwl. The AtomOwl feed in turn is serialised
as XHTML+RDFa; see listing 1.2 for an excerpt of the updates page.
Using AtomOwl over XHTML+RDFa allows both humans and machines to
consume the data updates properly. A human user directly accessing the page
is able to view the updates, a Semantic Web agent capable of understanding
XHTML+RDFa can process the feed entries for its purposes. A real-world ex-
ample of how to use the AtomOwl-feed is provided in the following. In this
experiment we have programmed SPARQLBot17 to access and query the Ato-
mOwl embedded in the riese updates page. SPARQLBot offers a Web-based
interface to define commands, which in turn maps to a SPARQL query (shown
in listing 1.3).
1 PREFIX aowl : < http :// bblfish . net / work / atom - owl /2006 -06 -06/# >
2 PREFIX rdf : < http :// www . w3 . org /1999/02/22 - rdf - syntax - ns # >
3
4 SELECT DISTINCT ? headline ? feed WHERE {
5 ? feed rdf : type aowl : Feed ;
6 aowl : entry ? entry .
7 ? entry aowl : title ? eTitle .
8 ? eTitle aowl : body ? headline .
9 }
10 LIMIT 10
Listing 1.3. A SPARQL query for data updates on riese.
17
http://semsol.org/semcamp/sparqlbot
Eventually, the same procedure can be applied to other scenarios, for exam-
ple, when attempting to consume news feeds in an online news-reader, such as
netvibes.com. It can be seen that certain indirections are necessary, however we
are confident that with the growing support of Semantic Web technologies—as
recently indicated by Yahoo!18 —the burdens are likely to vanish.
4 Towards Generalising User Contributed Interlinking
With the User Contributed Interlinking (UCI) we have proposed a novel ap-
proach for creating high-quality interlinks by relying on the users. The UCI ap-
proach is motivated by the observation that generic, template-based algorithms
(such as described in [7]) are limited regarding the quality of the typed links.
For large datasets such as riese where the entire European statistics are
brought to the Semantic Web it might appear impractical at first sight to man-
ually generate interlinks to other datasets. It is obvious that it is not feasible
to have one person dedicated to manually looking for adequate related sources.
However, by applying the Wiki-principle we want to initiate a crowdsourcing
process that encourages users to contribute to linked datasets with similar en-
thusiasm as they already show in the case of Wikipedia. It has to be noted that
the proposed UCI-feature is in an early stage of development and the first of its
kind. The current implementation as it can be found in riese is meant to boot-
strap the community-involvement in the area of linked datasets. It should be
adapted to other datasets as well. Based on the experiences gained with the first
release of UCI the system and the related processes will be refined. User accep-
tance is the critical success factor of UCI and therefore we aim at implementing
as many of the best practices of Wikipedia as possible.
Sanger [15] was actively involved in the beginning of Wikipedia and has
identified several factors that led to the great success of the platform such as
openness and ease of editing. By inviting everybody to contribute we clearly
highlight the openness of UCI. In addition we are working on enhancing the
user experience by constantly improving the user interface design and keeping
the user requirements at an absolute minimum as for instance no registration is
required for using the UCI.
One of the disadvantages of common Wikis as identified in [16] is the lim-
itation that “Wiki content is generally not available in a machine-processable
format”. With UCI we directly address this issue as the target outcome RDF is
machine-processable per se. There are nevertheless still challenges left, such as
reaching a critical mass of contributors by providing appropriate incentives or
addressing data provenance issues. However, we would like to see a conversion
of the strong community-engagement from Web 2.0 to the Semantic Web and
contribute to the initiation of this transformation by providing useful tools such
as the UCI.
As a next step, we have prototypically implemented a generalised UCI in a
demonstrator called irs (which is for interlinking of resources with semantics).
18
http://www.ysearchblog.com/archives/000527.html
Fig. 5. A demonstrator for a generalised UCI: irs.
The irs demonstrator—implemented with ARC19 —is available for testing pur-
poses at http://143.224.254.32/irs/. A screen shot of irs is shown in Fig. 5;
it enables users to create semantic links (currently owl:sameAs, rds:seeAlso,
and foaf:topic), to ask about existing links and to preview the (RDF) content.
Further, a simple version of provenance tracking is offered: By placing the state-
ments into a named graph (default is http://example.org/#unknown), one can
track down who stated what. A simple off-the-shelf SPARQL-endpoint is also
available in irs.
5 Discussion and Conclusion
While the Semantic Web itself may be regarded as a (backbone) infrastructure,
developers of Semantic Web applications have to be aware of issues arising with
it.
In this paper we have presented a Wiki-style approach for user contributed
(semantic) interlinking (UCI) in general, along with a discussion of tangible
19
http://arc.semsol.org/home
results. First we have implemented the UCI within riese, the RDFized and inter-
linked version of the European statistics. We have also addressed issues emerging
from using Semantic Web technologies in Web 2.0 (Ajax) environments.
With UCI we have showcased an approach potentially increasing the end-
user involvement in the Semantic Web. The acceptance of such features by the
community is crucial, hence we will keep working on improving the tools in order
to provide an enjoyable user experience.
Due to the usage of scripting languages, an efficient and effective development
of riese (and irs, alike) was made possible. The feature-testing cycle was kept
to a minimum; from a developer’s perspective it was possible to focus on the
important issue: functionality rather than configuration and heavy framework-
study.
Acknowledgements
The research reported in this paper was carried out in two projects: the “Knowl-
edge Space of semantic inference for automatic annotation and retrieval of multi-
media content” (K-Space) project20 , partially funded under the 6th Framework
Programme of the European Commission, and the “Understanding Advertising”
(UAd) project21 , funded by the Austrian FIT-IT Programme.
The authors would further like to thank the following people for their lively
input, discussions and support in implementation issues: Danny Ayers, Benjamin
Nowack, Tom Heath, and Richard Cyganiak.
References
1. G. Klyne, J. J. Carroll, and B. McBride. RDF/XML Syntax Specification (Re-
vised). http://www.w3.org/TR/rdf-concepts/, 2004.
2. C. Bizer, T. Heath, D. Ayers, and Y. Raimond. Interlinking Open Data on the
Web (Poster). In 4th European Semantic Web Conference (ESWC2007), pages
802–815, 2007.
3. G. Tummarello, R. Delbru, and E. Oren. Sindice.com: Weaving the Open Linked
Data. In The Semantic Web, 6th International Semantic Web Conference, 2nd
Asian Semantic Web Conference, ISWC 2007 + ASWC 2007, pages 552–565, 2007.
4. W. Halb, Y. Raimond, and M. Hausenblas. Building Linked Data For Both Humans
and Machines. In WWW 2008 Workshop: Linked Data on the Web (LDOW2008),
Beijing, China, 2008.
5. L. Sauermann, R. Cyganiak, and M. Völkel. Cool URIs for the Semantic Web. W3C
Editor’s Draft, W3C Semantic Web Education and Outreach Interest Group., 2007.
6. C. Bizer, R. Cyganiak, and T. Heath. How to Publish Linked Data on the Web.
http://sites.wiwiss.fu-berlin.de/suhl/bizer/pub/LinkedDataTutorial/,
2007.
20
http://kspace.qmul.net/
21
http://www.sembase.at/index.php/UAd
7. Y. Raimond, C. Sutton, and M. Sandler. Automatic Interlinking of Music Datasets
on the Semantic Web. In WWW 2008 Workshop: Linked Data on the Web
(LDOW2008), Beijing, China, 2008.
8. D. Brickley and R.V. Guha. RDF Vocabulary Description Language 1.0: RDF
Schema. W3C Recommendation, RDF Core Working Group, 2004.
9. B. Adida, M. Birbeck, S. McCarron, and S. Pemberton. RDFa in XHTML: Syn-
tax and Processing. W3C Working Draft 21 February 2008, W3C Semantic Web
Deployment Working Group, 2007.
10. D. Ayers. Graph Farming. IEEE Internet Computing, 12(1):80–83, 2008.
11. B. Adida. hGRDDL: Bridging microformats and RDFa. Web Semantics: Science,
Services and Agents on the World Wide Web, 6(1):54–60, 2008.
12. M. Hausenblas, W. Slany, and D. Ayers. A Performance and Scalability Metric
for Virtual RDF Graphs. In 3rd Workshop on Scripting for the Semantic Web
(SFSW07), Innsbruck, Austria, 2007.
13. M. Nottingham and R. Sayre. The Atom Syndication Format. RFC 4287, Network
Working Group, 2005.
14. D. Ayers and H. Story. AtomOwl Vocabulary Specification . Namespace Document,
Atom Owl Working Group, 2006.
15. L. Sanger. The Early History of Nupedia and Wikipedia: A Memoir. In C. DiBona,
M. Stone, and D. Cooper, editors, Open Sources 2.0: The Continuing Evolution.
O’Reilly, 2005.
16. D. E. O’Leary. Wikis: ’From Each According to His Knowledge’. Computer,
41(2):34–41, 2008.