NUI Galway – UL Alliance First Annual ENGINEERING AND - ARAN ...

Provenance in the Web of Data:
a building block for user profiling and trust in online communities
Fabrizio Orlandi, Alexandre Passant
Digital Enterprise Research Institute
National University of Ireland, Galway
fabrizio.orlandi@deri.org - alexandre.passant@deri.org
Abstract
Online collaborative knowledge bases such as
Wikipedia provide an extensive source of information,
not only to their readers, but also to a wide range of
applications and Web services. For example, DBpedia,
one of the largest datasets on the Web of Data, is
widely used as a reference for data interlinking and as
a basis for applications employing Semantic Web
technologies. Yet its dataset, directly derived from
Wikipedia articles, could contain errors due to
inexperience or anonymity of the contributors. By
analysing the Wikipedia edit history and the users'
contributions we provide detailed provenance
information for DBpedia statements and we make this
information publicly available on the Web of Data.
The dataset we provide is then fundamental for
analysing users' activities/interests and computing
trust measures.
Collaborative websites such as Wikipedia have
recently shown the benefit of being able to create and
manage very large public knowledge bases. However,
one of the most common concerns about these types of
information sources is the trustworthiness of their
content which can be arbitrarily edited by everyone.
The DBpedia project 1 , which aims at converting
Wikipedia content into structured knowledge, is then
not exempt from this concern. Especially considering
that one of the main objectives of DBpedia is to build a
dataset such that Semantic Web technologies can be
employed against it. Hence this allows not only to
formulate sophisticated queries against Wikipedia, but
also to link it to other datasets on the Web, or create
new applications or mashups. Thanks to its large
dataset and its cross-domain nature DBpedia has
become one of the most important and interlinked
datasets on the Web of Data. Therefore providing
information about where DBpedia data comes from and
how it was extracted and processed is crucial. This type
of information is called provenance and it describes the
entire data life cycle, from its origin to its subsequent
processing history.
Having provenance information about Wikipedia
data allows us to identify quality measures for
Wikipedia articles and estimate the trustworthiness of
their content. Then, since the DBpedia content is
directly extracted from Wikipedia, the same trust and
quality values can be propagated to the DBpedia
dataset. We apply this process to DBpedia, but this is
just one particular use-case, the same considerations
about provenance are suitable for every dataset on the
Web of Data. The benefits of using data provenance to
develop trust on the Web, and the Semantic Web in
particular, have been already widely described in the
state of the art. Provenance data provides useful
information such as timeliness and authorship of data.
It can be used as a ground basis for various
applications and use cases such as identifying trust
values for pages or pages fragments, or measuring
users' expertise by analysing their contributions and
then personalize trust metrics based on the user profile
of a person on a particular topic. Moreover, providing
also provenance meta-data as RDF and making it
available on the Web of Data offers more interchange
possibilities and transparency. This would let people
link to provenance information from other sources. It
provides them the opportunity to compare these sources
and choose the most appropriate one or the one with
higher quality. In our specific context of DBpedia for
example, by indicating by whom and when a RDF
triple was created (or contributed by), it could let any
application flag, reject or approve this statement based
on particular criteria.
In our work [1][2] we propose a modelling solution
to semantically represent information about provenance
of data in DBpedia and an extraction framework
capable of computing provenance for DBpedia
statements using Wikipedia edits. The framework
consists of: (i) a lightweight modelling solution to
semantically represent provenance of both DBpedia
resources and Wikipedia content, (ii) an information
extraction process and a provenance-computation
system combining Wikipedia articles' history with
DBpedia information, (iii) a set of scripts to make
provenance information about DBpedia statements
directly available when browsing this source, (iv) a
publicly available web service that exposes in RDF as
Linked Open Data our provenance dataset letting
software agents and developers consume it.
References
[1] Orlandi F., Champin P-A., Passant A., “Semantic
Representation of Provenance in Wikipedia,”, Semantic Web
Provenance Management workshop at ISWC2010, CEUR-
WS, Shanghai, 2010.
[2] Orlandi F., Passant A., “Modelling Provenance of
DBpedia Resources Using Wikipedia Contributions”,
Journal of Web Semantics, (to be published), 2011.
* This work has been funded in part by Science Foundation Ireland under Grant No. SFI/08/CE/I1380 (Lion-2) and by an IRCSET Scholarship.
1 http://dbpedia.org/
97