l Web Page Contents l Use top-k returned Web pages for each entity l Represent each set by a Virtual Document l Some heuristics § D (m): Top m (≤ k) documents are concatenated § T (all, n): Top n tokens with the highest weight from all top-k web pages § Snippet (m): Snippets of top m (≤ k) web pages § Probabilistic Language Model: KL-divergence l sim(ec, ei)  = doc_sim( vdoc(ec), vdoc(ei) )
			WIDM 2007