Comparing our results with some representative state-of-the-art systems.
| Bio-NER . | Method . | F-score . |
|---|---|---|
| (Okanohara et al. 2006) | Semi-Markov CRF + global features | 71.5 |
| (Hsu et al. 2009) | CRF + PSA(1) training | 69.4 |
| (Tsuruoka, Tsujii, and Ananiadou 2009) | CRF + SGD-L1 training | 71.6 |
| Our Method | CRF + ADF training | 72.3 |
| Segmentation | Method | F-score |
| (Gao et al. 2007) | Semi-Markov CRF | 97.2 |
| (Sun, Zhang, et al. 2009) | Latent-variable CRF | 97.3 |
| (Sun 2010) | Multiple segmenters + voting | 96.9 |
| Our Method | CRF + ADF training | 97.5 |
| Chunking | Method | F-score |
| (Kudo and Matsumoto 2001) | Combination of multiple SVM | 94.2 |
| (Vishwanathan et al. 2006) | CRF + SMD training | 93.6 |
| (Sun et al. 2008) | Latent-variable CRF | 94.3 |
| Our Method | CRF + ADF training | 94.5 |
| Bio-NER . | Method . | F-score . |
|---|---|---|
| (Okanohara et al. 2006) | Semi-Markov CRF + global features | 71.5 |
| (Hsu et al. 2009) | CRF + PSA(1) training | 69.4 |
| (Tsuruoka, Tsujii, and Ananiadou 2009) | CRF + SGD-L1 training | 71.6 |
| Our Method | CRF + ADF training | 72.3 |
| Segmentation | Method | F-score |
| (Gao et al. 2007) | Semi-Markov CRF | 97.2 |
| (Sun, Zhang, et al. 2009) | Latent-variable CRF | 97.3 |
| (Sun 2010) | Multiple segmenters + voting | 96.9 |
| Our Method | CRF + ADF training | 97.5 |
| Chunking | Method | F-score |
| (Kudo and Matsumoto 2001) | Combination of multiple SVM | 94.2 |
| (Vishwanathan et al. 2006) | CRF + SMD training | 93.6 |
| (Sun et al. 2008) | Latent-variable CRF | 94.3 |
| Our Method | CRF + ADF training | 94.5 |