Table 3 performance of our cnn-based and rpcnn-based systems for chemical-induced disease extraction
From: Chemical-induced disease extraction via recurrent piecewise convolutional neural networks
Methods | Without domain knowledge (%) | With domain knowledge (%) | ||||
|---|---|---|---|---|---|---|
P | R | F | P | R | F | |
CNN | 50.47 | 55.61 | 52.92 | 63.70 | 74.40 | 68.64 |
CNN + piecewise | 54.48 | 53.91 | 54.20 | 63.83 | 75.16 | 69.03 |
CNN + attention | 48.40 | 58.54 | 52.99 | 62.28 | 76.58 | 68.69 |
CNN + attention+piecewise | 57.80 | 54.20 | 55.94 | 59.97 | 81.49 | 69.09 |
RPCNN | 55.17 | 63.63 | 59.10 | 65.24 | 77.21 | 70.77 |