Table 4.
20 | HMM-3M | 3!-GRM | 3!-GRM/HMM-3M |
Recall | 0.446 | 0.394 | 0.449 |
Precision | 0.701 | 0.583 | 0.710 |
100 | HMM-3M | 3!-GRM | 3!-GRM/HMM-3M |
Recall | 0.687 | 0.646 | 0.689 |
Precision | 0.404 | 0.364 | 0.412 |
1000 | HMM-3M | 3!-GRM | 3!-GRM/HMM-3M |
Recall | 0.911 | 0.917 | 0.917 |
Precision | 0.102 | 0.104 | 0.104 |
A ranking is first produced by 3!-GRM and the top 1000 ranks are then re-scored by HMM-3M to produced the results reported. For ease of comparison, the results of HMM-3M and n!-GRM methods from Tables 3 and 4 are also included.