Databases Reference
In-Depth Information
5
Web 2.0 Approaches
Besides interactive desktop tools, researchers have started to explore how to use
communities of users to develop ontology matchings collaboratively and to share
them.
Crowdsourcing
- outsourcing of a task to a community of motivated individ-
uals - has had huge success in projects such as Wikipedia and social bookmarking
sites such as Digg. Similar wisdom of the crowd approaches are beginning to gain
traction in the matching community.
Zhdanova and Shvaiko developed an online application to support and collect
community-driven matchings [
Zhdanova 2005
]. The web application allowed users
to upload ontologies and to use online tools to perform an automatic matching
between the ontologies. Once the users generated the matching, they could save and
share it with other members of the community. Realizing that matchings can often
be subjective, the authors designed their application to collect information about the
users of the community in terms of their expertise, experience levels with particular
ontologies, and their goals for a particular matching. Other members of the com-
munity could therefore make informed decisions about whether or not to rely on an
uploaded matching. The application also stored information about the relationship
between users of the community.
Similarly, the OntoMediate Project, as part of their research initiative, has been
exploring to what extent collaborative online environments can help to facilitate the
specification of ontology matchings [
Correndo et al. 2008b
]. The prototype system
supports the matching of local ontologies to already uploaded ontologies and match-
ings. Furthermore, the automated procedures make use of the existing matchings to
improve the quality of suggested matchings. The tools exploit social interaction to
help improve matching quality. Users of the community that work with similar data
can socially interact with each other to help validate matchings, spot errors, provide
feedback, and propose alternatives [
Correndo et al. 2008a
].
McCann et al. have also been exploring Web 2.0 approaches. They have proposed
an interesting approach to engage the user community [
Robert McCann et al. 2008
].
In their research, they have been investigating how to gather feedback from users in
the form of simple questions in which the answers are used to improve the accuracy
of the underlying algorithms. The goal is to pose questions to users that will have a
significant impact on the tool's accuracy, as well as be questions that are easy for a
human to answer but difficult for a machine. For example, an automated procedure
may guess that a particular attribute is of type date, but may not be completely confi-
dent about the choice. User-expertise can be exploited in this circumstance to clarify
whether the particular attribute is a date or not, leading to significant improvement
in the algorithm choices.
In BioPortal,
9
an online tool for accessing and sharing biomedical ontologies,
researchers have been exploring the impact of supporting matchings as a form of
ontology metadata. Users can upload matchings that are generated offline as well as
9
http://bioportal.bioontology.org/
.
