BDBComp
Parceria:
SBC
Semantic-Based Query Reformulation for PDMS

Damires SouzaAna Carolina Salgado

A Peer Data Management System (PDMS) is a P2P application which enables users to transparently query several heterogeneous and autonomous data sources. Semantic PEEr-to-Peer Data Management System - SPEED is a semantic-based PDMS which adopts a mixed architecture (DHT and super peer). In this architecture, data peers (associated to data sources) are grouped according to their domain forming one or more semantic clusters. Every semantic cluster has a special type of peer, called integration peer, which is responsible for complex processes like query processing. When a query arrives at the integration peer, it searches the query index and looks at semantic mappings to find out peers which are able to answer the query. Then, it reformulates the query into a set of queries which will be sent to the data peers and other integration peers. The query reformulation is done according to semantic mappings. Answers from every peer are returned to the integration peer, where they are integrated to produce the query result. The final result is sent back to the data peer from which the query was submitted. However, the key step in query processing is reformulating a peer's query over other peers on the available semantic mapping paths. In this sense, this work aims to develop a semantic-based query reformulation process for the SPEED system. In order to develop such process, two relevant issues will be dealt: semantic mapping definition and semantic knowledge application. To the former, we will formalize a semantic mapping language which will be used to define mappings and their usage. To the latter, we use ontologies, metadata and context information, i.e. the circumstantial information that makes a situation unique and comprehensible, as a way to enrich the overall query processing and the query reformulation step as well.

http://www.lbd.dcc.ufmg.br:8080/colecoes/wtdbd/2007/012.pdf

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato: bdbcomp@lbd.dcc.ufmg.br
     Mantida por:
LBD