Table 4 The performance of protein coreference resolution with different models on two evaluation datasets of BioNLP
From: Distinguished representation of identical mentions in bio-entity coreference resolution
Dev | Test | |||||
|---|---|---|---|---|---|---|
P | R | F1 | P | R | F1 | |
Hybrid | 59.9 | 77.1 | 67.4 | 55.6 | 67.2 | 60.9 |
Simple | 63.4 | 64.4 | 63.9 | 46.3 | 50.0 | 48.1 |
Bio-SCoRes | 72.4 | 63.2 | 67.5 | / | / | / |
Lee2018-rule | 68.8 | 76.0 | 72.2 | 60.2 | 63.8 | 62.0 |
lee2018-neural | 60.4 | 61.9 | 61.2 | 54.9 | 58.0 | 56.4 |
Bioe2e | 71.7 | 56.7 | 63.1 | 55.6 | 47.5 | 51.2 |
BioNeu | 77.1 | 61.9 | 68.7 | 71.5 | 60.9 | 65.8 |
BioNeu-feature | 75.5 | 65.8 | 70.4 | 69.5 | 60.2 | 64.5 |
BioNeu + SFA | 73.0 | 65.3 | 69.0 | 72.3 | 61.6 | 66.5 |
KE-LSTM | 68.1 | 63.4 | 65.6 | 69.6 | 69.4 | 69.5 |
KE-LSTM-feature | 74.4 | 64.8 | 69.3 | 62.8 | 61.2 | 62.0 |
KE-LSTM + SFA | 70.8 | 68.3 | 69.6 | 69.5 | 68.2 | 68.8 |