Table 7 The detailed performance of MPANN on different batches of the BioASQ test set
From: Multi-probe attention neural network for COVID-19 semantic indexing
Batch | #Articles | EBP(%) | EBR(%) | EBF(%) | MaP(%) | MaR(%) | MaF(%) | MiP(%) | MiR(%) | MiF(%) |
|---|---|---|---|---|---|---|---|---|---|---|
Test batch 1, week 1 | 7,967 | 73.36 | 60.56 | 64.78 | 68.88 | 53.52 | 54.15 | 73.33 | 59.41 | 65.64 |
Test batch 1, week 2 | 10,053 | 71.32 | 58.97 | 62.76 | 67.72 | 53.04 | 53.65 | 71.06 | 57.58 | 63.61 |
Test batch 1, week 3 | 4,870 | 72.04 | 60.31 | 64.05 | 67.49 | 52.43 | 52.86 | 71.92 | 59.06 | 64.86 |
Test batch 1, week 4 | 5,758 | 71.81 | 58.79 | 62.94 | 68.15 | 52.29 | 53.08 | 71.87 | 57.52 | 63.90 |
Test batch 1, week 5 | 5,770 | 70.81 | 60.34 | 63.48 | 66.77 | 52.13 | 52.25 | 70.43 | 59.41 | 64.45 |
Test batch 2, week 1 | 6,376 | 72.11 | 58.15 | 62.60 | 68.40 | 51.11 | 52.02 | 72.04 | 57.00 | 63.64 |
Test batch 2, week 2 | 9,101 | 70.93 | 58.29 | 62.31 | 67.74 | 53.08 | 53.63 | 70.99 | 57.48 | 63.52 |
Test batch 2, week 3 | 7,013 | 71.88 | 58.30 | 62.81 | 68.05 | 52.74 | 53.41 | 71.81 | 57.25 | 63.71 |
Test batch 2, week 4 | 6,070 | 73.35 | 59.64 | 64.17 | 69.31 | 53.35 | 53.93 | 73.04 | 58.51 | 64.98 |
Test batch 2, week 5 | 6,151 | 73.14 | 61.17 | 64.94 | 68.70 | 53.02 | 53.76 | 73.16 | 59.84 | 65.83 |
Test batch 3, week 1 | 5,890 | 73.31 | 60.09 | 64.59 | 68.69 | 51.59 | 52.46 | 73.31 | 59.20 | 65.51 |
Test batch 3, week 2 | 10,818 | 72.61 | 59.10 | 63.55 | 68.27 | 53.05 | 53.97 | 72.38 | 57.84 | 64.30 |
Test batch 3, week 3 | 4,022 | 71.43 | 60.79 | 64.07 | 68.46 | 50.86 | 51.38 | 71.61 | 59.50 | 65.00 |
Test batch 3, week 4 | 5,373 | 72.89 | 60.93 | 64.80 | 68.25 | 52.82 | 53.35 | 72.62 | 59.45 | 65.38 |
Test batch 3, week 5 | 5,325 | 70.93 | 60.53 | 63.63 | 67.08 | 54.26 | 54.17 | 70.70 | 59.36 | 64.54 |