Figure 5

Coverage of the cluster-by-topic list across a range of queries. Anonymous queries in the Anne O'Tate query web log were analyzed. For each query, the coverage was computed (i.e., the proportion of MeSH-indexed articles in the PubMed search output that were included in the 15 MeSH-based topical clusters). The results were averaged for retrieved literatures of different size ranges as follows: 0–100 articles, 6 queries; 101–1000 articles, 9 queries; 1001–10000 articles, 9 queries; and >10000 articles, 3 queries.