Alvis - Superpeer Semantic Search Engine (2004-2006)
ALVIS was a European STREP project in the Semantic-based knowledge
systems program that has built a semantic-based peer-to-peer search
engine. A consortium
of eleven partners from six different European Community countries plus Switzerland and
China contributed expertise in a broad range of specialities including network topologies, routing
algorithms, probabilistic approaches to information retrieval, linguistic analysis and bioinformatics.
EPFL worked on the design and evaluation of a peer-to-peer full-text
- We have designed a communication protocol between peers and superpeers.
- We have performed a scalability analysis which demonstrates that our distributed indexing cost grows linearly with the number of documents.
- We have designed distributed algorithms for indexing and content-based ranking in structured P2P overlay, and integrated an efficient implementation for maintaining global document collection statistics.
We can conclude that our approach shows promise to scale to very
large scale peer populations.
We believe the proposed model and associated prototype convincingly demonstrate that P2P
web-scale retrieval is feasible and hope that our work will contribute to the progress in a domain
that is generally recognized as crucial for the development of less centralized and therefore more
user-centered information dissemination and management techniques.