Table 3 Log of Probability Measures for test set optimized single parameters. The best performance in each row is marked with an asterisk.
From: Improving a gold standard: treating human relevance judgments of MEDLINE document pairs
Judge | M 1 | M 4 | M 5 |
|---|---|---|---|
0 | -8902 | -8425 | -7805* |
1 | -7087 | -6925 | -6833* |
2 | -6872 | -6641* | -6674 |
3 | -6760 | -6675 | -6462* |
4 | -7694* | -8068 | -7703 |
5 | -7121 | -7259 | -6942* |
6 | -7032 | -7015 | -6913* |
7 | -7094 | -6482* | -6836 |
8 | -7352 | -6487* | -7199 |
9 | -7113 | -6992 | -6874* |
10 | -8041 | -7576 | -7442* |
11 | -7275 | -7450 | -6909* |
12 | -8160 | -7694* | -7784 |
Ave | -7423 | -7207 | -7106 |