Presentation on theme: "Measuring the Semantic Web Rosa Gil Iranzo GRIHO, Universitat de Lleida, Spain Roberto García González rhizomik.net."— Presentation transcript:
Measuring the Semantic Web Rosa Gil Iranzo GRIHO, Universitat de Lleida, Spain Roberto García González rhizomik.net
Outline Motivation why to measure? Approach complex systems Measuring applying statistical tools Results is the semantic web a complex system? Conclusions
Motivation Semantic Web, an open evolving system. TimBL: Looking for a metric in The Fractal nature of the Web, Design Issues. How is it measured? Whats the metric?
Motivation Why to measure? From the TimBLWeaving the Web Semantic Web plan… –Where we are now? –How is it evolving? –Are we going where it was planned? –…
Approach Semantic Web as complex as many other systems: –metabolic networks –acquaintance networks –food webs –neural networks –The WWW –…
Approach This complex systems are studied using Complex Systems (CS) Analysis. Statistical tools for graph models: –Degree Distribution –Small World –Clustering Coefficient –…
Approach Model the system as a graph. CS graph characteristics: –Degree Distribution power law, P(k) k - r –Small World small diameter, d d random –Clustering Coefficient high clustering, C >> C random
Measuring Is the Semantic Web a CS? It is already a graph. Crawl all DAML Ontologies Library: –2003: 56,592 nodes, 131,130 arcs –2005: 307,231 nodes and 588,890 arcs Statistical study of the graph.
Results NetworkNodes C DAMLOntos (2003-4-11) 56,5924.630.1524.37-1.48 DAMLOntos (2005-1-31) 307,2313.830.0925.07-1.19 WWW ~200 M 0.1083.10 -2.24 WordNet66,025 0.0607.40-2.35 WordsNetwork500,000 0.6872.63-1.50
Results It is a small world diameter smaller than random graph, d=4.37 while d rand =7.23 It has high clustering C=0.152 while C random =0.0000895 It is scale free power law degree distribution, P(k)k –1.19
Results CDF (Cumulative Distribution Function) Degree
Conclusions The Semantic Web is a Complex System. Behaves like a living system (neural network, food web, proteins net,…), i.e. the same dynamics. Same behaviour 2003-2005.
Conclusions Just exploring applications: –Degree dynamics for trust computation. –Ontology alignment (clusters, centrality,…). –Metadata high volumes management. – etc. More information and tools at: http://rhizomik.net/livingsw http://rhizomik.net/livingsw
Thank you for your attention Roberto García Rosa Gil