Table 6 Experimental Results
From: An improved data augmentation approach and its application in medical named entity recognition
Models | Precision(%) | Recall(%) | F1(%) |
|---|---|---|---|
â‘ Word2Vec-BiLSTM-CRF | 80.000 | 77.907 | 78.940 |
②Word2Vec-BiLSTM-CRF (CRR) | 80.744 | 77.757 | 79.223 (+ 0.28) |
③Word2Vec-BiLSTM-CRF (TER) | 80.120 | 78.091 | 79.092 (+ 0.15) |
â‘£BERT-CRF | 83.133 | 81.956 | 82.540 |
⑤BERT-CRF(CRR) | 82.474 | 83.422 | 82.945 (+ 0.41) |
⑥BERT-CRF (TER) | 82.640 | 83.039 | 82.839 (+ 0.30) |
⑦BERT-BiLSTM-CRF | 80.338 | 83.939 | 82.099 |
⑧BERT-BiLSTM-CRF (CRR) | 82.122 | 85.105 | 83.587 (+ 1.49) |
⑨BERT-BiLSTM-CRF (TER) | 83.328 | 82.606 | 82.965 (+ 0.87) |
â‘©RoBERTa-wwm-ext-CRF | 82.042 | 83.122 | 82.579 |
⑪RoBERTa-wwm-ext-CRF (CRR) | 82.323 | 83.256 | 82.787 (+ 0.21) |
⑫RoBERTa-wwm-ext-CRF (TER) | 81.734 | 83.722 | 82.716 (+ 0.14) |
⑬RoBERTa-wwm-ext-BiLSTM-CRF | 82.788 | 83.505 | 83.145 |
â‘RoBERTa-wwm-ext-BiLSTM-CRF (CRR) | 82.756 | 83.639 | 83.195 (+ 0.05) |
⑮RoBERTa-wwm-ext-BiLSTM-CRF (TER) | 83.276 | 84.872 | 84.066 (+ 0.92) |