Capturing, Representing and Operationalising Semantic Integration (CROSI) project - final report

Kalfoglou, Y., Hu, B., Reynolds, D. and Shadbolt, N. (2005) Capturing, Representing and Operationalising Semantic Integration (CROSI) project - final report. Technical Report , ECS, University of Southampton.

Download

Preview

PDF
616Kb

Abstract

n October of 2004, the $\underline{University of Southampton}$ and
$\underline{Hewlett Packard Laboratories at Bristol}$ joined forces to
work on a collaborated project, CROSI, in order to investigate
semantic integration. The project lasted 12 months and its primary
focus was the use of Artificial Intelligence technology, like
ontologies, in the Semantic Web environment in order to research
issues related to ontology mapping.

A number of new insights and emergent issues were identified by
using a practical testbed and a thorough evaluation strategy. The
project's main deliverables were: (a) a $\emph{comprehensive survey}$
of state-of-the-art technologies that address semantic integration;
(b) engineering artifacts for guiding the potential semantic
integration practitioner to develop a semantic integration system:
the $\emph{Semantic Intensity Spectrum}$ (SIS) and a $\emph{Modular Architecture}$ for developing semantic integration systems; (c) an
ontology mapping prototype system, $\emph{CMS (CROSI Mapping System)}$
which acts as a demonstrator of mapping techniques.

CROSI also shed light and provided fruitful insights for potential
extensions of this technology. For example, it was proven that
combining different matchers gives us a better alignment than using
them independently. But, we also identified areas where more
research is needed, like for example, practical and re-usable
algorithms for aggregating alignment results.

The project also had a good visibility record and achieved a decent
dissemination across various communities (Web community, Knowledge
Engineering community), especially when its short life cycle is
taken into account. The highlight was our participation in the
annual ontology alignment contest where CMS was ranked in the top
three systems for alignment and it was one of the two out of seven
participating systems that managed to parse correctly and without a
flaw the industrial-strength, sizeable ontologies in the medical
informatics domain.

In this final project report, we elaborate on the issues mentioned
above. This report is also accompanied by a software system and its
manual where more technical details regarding the mechanisms used
for alignment are explained.