Information Technology Reference
In-Depth Information
network resulting in redundant query messages in
the network. Cai, et al. (2004) proposed a scalable
and distributed RDF repository called RDFPeers
based on a structured P2P system. RDFPeers or-
ganize into a multi-attribute addressable network
(MAAN) (Cai, et al., 2003) which extends Chord
to efficiently answer multi-attribute and range
queries. When an RDF triple is inserted into the
network, it will be stored three times by applying
a globally-known hash function to its subject,
predicate, and object. We take a similar approach
to deploy Chord as the substrate for the lower-tier
network, however, we store the <sub pred> and
<pred obj> pairs for each data triple as the predi-
cate is always known in a context query. Thus, the
cost of inserting RDF triples into the network can
be reduced. In addition, the identifier space of the
lower-tier in our network is much smaller than the
one in RDFPeers. Hence, the maintenance cost is
lower as compared to RDFPeers since each peer
maintains fewer neighbors. Tang, et al. (2003)
applied classical Information Retrieval techniques
to P2P systems and built a decentralized P2P
information retrieval system called pSearch. The
system makes use of a variant of CAN to build
the semantic overlay and uses Latent Semantic
Indexing (LSI) (Deerwester, et al., 1990) to map
documents into term vectors in the space. Li, et
al. (2004) built a semantic small world network in
which peers are clustered based on term vectors
computed using LSI. They proposed an adaptive
space linearization technique for constructing link
structures. While we take the semantic approach
which is conceptually similar to (Tang, et al.,
2003) and (Li, et al., 2004), we propose the use of
schema-based metadata to extract data semantics.
The formal design of ontologies minimizes the
problems of synonyms and polysemy incurred by
VSM, and incurs a lower overhead than LSI does.
Kleinberg (Kleinberg, 2000) proposed the small
world network model where every node maintains
four links to each of its closest neighbors and one
long distance link to a node chosen from a prob-
ability function. He has shown that a query can
be routed to any node in O(log 2 n) hops, where n
is the total number of nodes in the network. We
build the upper-tier network based on the small
world network model. The small world model has
many advantages, such as it is easy to construct
and the number of state information that each
node maintains is fixed and not proportional to
the number of semantic clusters. In our earlier
work (Gu, et al., 2005), we have proposed a se-
mantic P2P network for context search by using a
Gnutella-like network as the substrate. However,
the flooding-based routing mechanism is not very
efficient in terms of search path and scalability.
This article proposes a more efficient and scal-
able semantic network based on a structured P2P
network (i.e., Chord).
CONCLUSION
In this article, we present an ontology-based
semantic P2P network for searching context
information in wide-area networks. The prelimi-
nary results have shown that our system has good
search efficiency and low cost of node joining
and leaving, and our system can scale to a large
number of peers. The use of our system is not
limited to the context-aware computing domain;
in fact, it applies to any P2P searching system
where schemas are explicitly defined.
REFERENCES
Cai, M., & Frank, M. (2004). RDFPeers: A Scal-
able Distributed RDF Repository based on A
Structured Peer-to-Peer Network . Paper presented
at the Proceedings of the 13th International World
Wide Web Conference, New York.
Cai, M., Frank, M., Chen, J., & Szekely, P. (2003).
MAAN: A Multi-attribute Addressable Network for
Grid Information Services . Paper presented at the
Proceedings of the 4th International Workshop on
Grid Computing.
Search WWH ::




Custom Search