Information Technology Reference
In-Depth Information
cal data repository which supports RDF-based
query using RDQL (RDQL). Upon creation, each
producer will first go through the ontology-based
semantic mapping process to extract the seman-
tics of its local data. It will then join a semantic
cluster by applying the SHA1 hash function to
the semantics of its main data. These semantic
clusters logically form the upper-tier network in
which each node builds its routing index based
on the small world network model (Kleinberg,
2000). In the lower-tier network, nodes in each
semantic cluster are organized as Chord for storing
context data and routing context queries in a loga-
rithmic number of hops. Upon receiving a context
query, the node first pre-processes it to obtain the
semantic cluster associated with the query, and
then routes it to an appropriate semantic cluster.
In the lower-tier, the node routes the query using
its finger table. Nodes that receive the query do
a local search, and return results.
used as semantic clusters, and denoted as set E =
{Service, Application, Device, ...} . The mapping
computation is done locally at each peer. For the
mapping of RDF data, a peer needs to define a
set of lower ontologies and store them locally.
Upon joining the network, a peer first obtains the
upper ontology and merges it with its local lower
ontologies. Then it creates instances (i.e., RDF
data) and adds them into the merged ontology
to form its local knowledge base. A peer's local
data may be mapped into one or more semantic
clusters by extracting the subject, predicate and
object of an RDF data triple. Let SCn sub , SCn pred
and SCn obj where n = 1, 2, ... denote the semantic
clusters extracted from the subject, predicate and
object of a data triple respectively. Unknown
subjects/objects (which are not defined in the
merged ontology) or variables are mapped to E .
If the predicate of a data triple is of type Object-
Property , we obtain the semantic clusters using
( SC1 pred SC2 pred ⋃ ... SCn pred ) ⋂ ( SC1 obj SC2 obj
⋃ ... SCn obj ). If the predicate of a data triple is of
type DatatypeProperty , we obtain the semantic
clusters using ( SC1 sub SC2 sub ⋃ ... SCn sub ) ⋂
( SC1 pred SC2 pred ⋃ ... SCn pred ). Examples 1 and
2 in Figure 2a show the RDF data triples about
the location and light level in a bedroom provided
by a producer peer. In Example 2, we first obtain
the semantic clusters from both the subject and
predicate, and then intersect their results to get
the final semantic cluster - IndoorSpace .
A context query follows the same procedure
to obtain its semantic cluster(s), but it needs all
the sets of lower ontologies. In real applications,
users may create duplicate properties in their
lower ontologies which conflict with the ones in
the upper ontology. For example, the upper ontol-
ogy defines the rdfs:range of predicate locatedIn
as Location whereas the lower ontology defines
its rdfs:range as IndoorSpace . To resolve this
issue, we create two merged ontologies, one for
clustering peers and the other for clustering que-
ries. If such a conflict occurs, we select the af-
fected properties defined in the lower ontology
ONTOLOGY-BASED
SEMANTIC CLUSTERING
In this section, we describe how to use ontology-
based metadata to extract the semantics of both
RDF data and queries, and map them into appro-
priate semantic clusters. In our system, context
data are described as RDF triples based on a set
of context ontologies. We adopt a two-level hi-
erarchy in the design of context ontologies. The
upper ontology defines common concepts in a
computing domain, e.g., context-aware comput-
ing, and it is shared by all peers. Each peer can
define its own concepts in its lower ontologies.
Different peers may store different sets of lower
ontologies based on their application needs. The
upper ontology can be extended with new concepts
and properties upon the agreement among all the
peers in the network.
To illustrate the semantic mapping process, we
use an example of ontology as shown in Figure
1. All the leaf nodes in the upper ontology are
Search WWH ::




Custom Search