Table 5 Comparison of training time between CLSTM (contextual long short-term memory networks [LSTMs] with conditional random fields [CRF])and comparative methods for the NCBI, GM, and CDR corpora
From: Biomedical named entity recognition using deep neural networks with contextual information
Training time (Hours) | NCBI | GM | CDR | |
|---|---|---|---|---|
BiLSTM | 4.08 | 10.77 | 4.39 | |
BiLSTM-CRF | 4.70 | 12.56 | 5.23 | |
GRAM-CNN | 11.27 | 34.08 | 12.64 | |
BERT | 1.01 | 10.04 | 3.72 | |
CLSTM | word level | 5.54 | 14.57 | 5.98 |
| Â | character level | 4.84 | 13.30 | 5.64 |
| Â | word+char levels | 5.84 | 14.73 | 5.91 |
| Â | Average | 5.41 | 14.20 | 5.84 |