EXPERIMENTAL RESULTS
We have implemented in Java our proposal using JENA4 for managing and storing RDF metadata, and ARQ5 for querying
these metadata with the SPARQL query language. Moreover, in order to be as much general as possible, we generate randomly
RDF graphs by specifying particular graph sizes and densities using RDFizer6 (especially, randomRDF). In our proposal,
multimedia content URIs are linked to remote server URIs (cf., Section III-B), in order to preserve this information for each
randomly generated RDF description we select randomly a specific RDF node and typed it as a server.
Thereafter, we evaluate the efficiency of the metadata resume by measuring the response time of answering to user queries.
For particular contexts, i.e., specific number of servers and RDF graph sizes, we propose to:
Run an initial query on a centralized metadata resume which has not been summarized (i.e., containing all metadata) and
on one remote server metadata collection.
Run a relaxed query on a centralized metadata resume which has not been summarized and on another centralized metadata
resume which have been summarized.
Run a relaxed query which seeks for servers that may contain the desired information on a centralized metadata resume.