Component | document | aspect | passage2 |
|---|
Best of baselines | 0.2906 | 0.2189 | 0.0988 |
UniNE1+York07ga2+kyoto1 | 0.2743 (-5.60%) | 0.2065 (-5.63%) | 0.0978 (-1.01%) |
UniNE1+York07ga2+UBexp1 | 0.2802 (-3.56%) | 0.2219 (1.38%) | 0.1047 (5.96%) |
UniNE1+MuMshFd+kyoto1 | 0.2828 (-2.66%) | 0.2221 (1.46%) | 0.0997 (0.86%) |
UniNE1+MuMshFd+UBexp1
| 0.2906 (0.00%) |
0.2380 (8.75%)
|
0.1059 (7.19%)
|
- Corresponding to the baselines, we evaluate the combinations using the reciprocal method in this table. In total, there are four combinations generated from three different IR models. Each combination contains a DFR baseline, a BM25 baseline and a LM baseline. The values in the parentheses are the relative rates of improvement over the best results of the baselines. One of the conclusions is that the alliance of giants with boldface is the winner on all the measures.
