Table 2.

Comparisons of different term ranking algorithms for different term reweighting methods (R = 10 and E = 25)

The mean average precision is presented for a combination of term ranking (rows) and term reweighting (columns) methods, including, in parentheses, the percent (%) improvement from the unexpanded queries.

EMIM: expected mutual information measure, F4: F4MODIFIED, IDF: inverse document frequency, KLD: Knullback-Leibler divergence, LCA: local context analysis, RSV: Robertson selection value.

^a(p < 0.05) and ^b(p < 0.01) are in bold.