By exploiting open distributed architectures like Peer-to-Peer (P2P) and Grids, networked organizations can dynamically cooperate and share resources often in response to opportunities or challenges that cannot be anticipated in advance and require a rapid response [1, 7]. One of the major obstacles in the deployment of networked organizations like Virtual and Extended Enterprises is the lack of semantic interoperability techniques, to adequately ensure the effective, coordinated, and virtualized access to distributed and heterogeneous resources, under dynamic and context-dependent requirements. In particular, in order to support the correct and effective sharing and composition of informational resources to provide complex services in virtual organizations, an important requirement is the availability of a semantic description of the resources to be shared, to make them understandable and usable by the target community. In addition, to enable seamless access and retrieval of the right information resources, in the time frame that the users require, while preserving the information representation and management requirements of each single party involved in the networked organization coalition, appropriate matchmaking techniques are required, to dynamically perform the matching of requests for information resources against a multitude of, possibly heterogeneous, ontology descriptions. In this paper, we address the problem of semantic interoperability in networked, multi-ontology enterprise contexts, where ontologies are employed for representing informational resources that each party provides for dynamic sharing and exchange with other parties. We study the problem of semantic interoperability under the assumption that no agreement exists among the various parties about the way resources are described in each ontology and about the vocabulary to be adopted for semantic description of resources. We propose a reference ontology model and associated semantic matching techniques to enforce dynamic resource discovery, by exploiting semantic resource descriptions available in an ontology ? This paper has been funded by NoE INTEROP, IST Project n. 508011 - 6th EU Framework Programme and by “Wide-scalE, Broadband, MIddleware for Network Distributed Services (WEB-MINDS)” FIRB Project funded by the Italian Ministry of Education, University, and Research.
at different levels of detail and with different degrees of flexibility. Then, we discuss ongoing and future research work on exploiting such techniques for dynamic semantic collaborations in networked organizations based on the P2P paradigm. 2
To support dynamic resource discovery in networked multi-ontology contexts, we have developed a reference ontology model and associated ontology matching techniques to dynamically identify semantic mappings among similar concepts in different ontologies.
Reference ontology model. The advent of the Semantic Web has produced
a large body of research around ontology languages, and many standard
proposals that can be used for resource description in open networked systems have
emerged (e.g., RDF, DAML+OIL, OWL). Different ontologies can describe the
same domain using different descriptions of the same resources, also using the
same language (e.g., in OWL ontologies, the same real world resources can be
described in many different forms, due to the syntactical freedom of RDF). An
important requirement of ontology matching techniques is to capture the
elements that are relevant for matching purposes in ontology resource descriptions
in a language-independent manner, to be applicable in many scenarios. To this
end, we have defined a reference ontology model, called H-Model, capable of
representing the ontology features that are relevant for matching purposes in a
language independent way, in terms of concepts, properties, and semantic
relations. A node ontology is organized as a two-layer architecture where the upper
layer represents the content knowledge and the lower layer represents the network
knowledge, respectively. The content knowledge layer describes the knowledge of
a node, namely the knowledge about the resources that a node brings to the
networked organization. The network knowledge layer provides a description of
other parties of the networked organization providing resources that are
semantically related to those described in the content knowledge layer of the considered
node. A detailed description of the reference ontology model and of the
architecture of a node ontology can be found in [
Ontology matchmaking techniques. In open networked contexts, nodes
dynamically request to other parties for resources semantically compatible with
a target resource (dynamic resource discovery). In this context, each node has
to compare incoming requests against its node ontology, in order to discover
whether it can provide resources matching the target. In order to address this
requirement, appropriate ontology matching techniques are required capable to
cope with different levels of detail in concept descriptions. For ontology
matching, we note that in a node ontology, the meaning of ontology elements depends
basically on the names chosen for their definition and on the relations they have
with other elements in the ontology. An important requirement is related to
the fact that these features can have a different impact in different ontology
structures. We have developed matching techniques capable of performing
dynamic matching at different levels of depth, with different degrees of flexibility
and accuracy of results by taking into account various metadata elements (i.e.,
concept name, concept properties, concept relations) separately or in
combination. Different matching models have been conceived with the goal of providing
a wide spectrum of metrics suited for dealing with many different matching
scenarios that can be encountered in comparing ontological resource descriptions
in real open distributed systems. The choice of the appropriate matching model
depends on the level of detail of the ontology description as well as on the
expected degree of precision of the results. The shallow model is useful when only
concept names are to be considered. It requires few computational resources
since neither concepts properties nor their relations are considered. This model
is well suited, for example, to perform an initial ontology comparison to decide
wether it is worth performing a deeper analysis. If the ontology is constituted
mainly by concepts with a few number of properties and hierarchical relation
among concepts, the shallow and deep model allow a good degree of precision
without requiring great amount of computational resources. In presence of an
articulated ontology, with rich resource descriptions and where relations among
concepts are described through property values, the intensive model guarantees
the most accurate results, but is the most expensive in term of computational
resources. A description of the H-Match algorithm and of its matching models
is given in [
We are testing the H-Match algorithm on real matching cases, to evaluate and
experiment performance and scalability issues posed by ontology-based query
resolution considering large ontologies (e.g., Semantic Web ontologies).
Furthermore, we will extend our matchmaking techniques to take into account also
aspects related to the quality of ontology information. Future research activity
will address the application and extension of the ontology-based semantic
interoperability methods and techniques presented in this paper in the framework
of the Helios (Helios Evolving Interaction-based Ontology knowledge Sharing)
system conceived for supporting P2P-based semantic collaboration [
In Helios, we are working in the direction of defining a semantic-based query
distribution protocol in order to support an effective knowledge discovery
process and the spontaneous and dynamic definition of semantic-based communities
of interest [