Table 6 Experimental results of all models on AgCNER dataset.
Model | Deep Learning | External Features | LLM | Precision | Recall | F1 |
|---|---|---|---|---|---|---|
HMM | ✗ | ✗ | ✗ | 63.58 | 72.02 | 67.54 |
CRF | ✗ | ✗ | ✗ | 94.16 | 92.2 | 93.17 |
BiLSTM-CRF | ✓ | ✗ | ✗ | 93.26 | 93.9 | 93.58 |
BiLSTM-Attention-CRF | ✓ | ✗ | ✗ | 94.04 | 93.04 | 93.53 |
IDCNN-CRF | ✓ | ✗ | ✗ | 92.38 | 93.37 | 92.88 |
Lattice-LSTM | ✓ | Character-word Lattice | ✗ | 90.4 | 93.13 | 91.74 |
TENER | ✓ | Character and Word-level Embeddings | ✗ | 92.79 | 94.95 | 93.85 |
FLAT | ✓ | Flat Lattice | ✗ | 93.65 | 94.87 | 94.26 |
NFLAT | ✓ | Lexicon | ✗ | 94.21 | 95.10 | 94.66 |
Graph4CNER48 | ✓ | Lexicon, Collaborative Graph Network | ✗ | 92.84 | 93.59 | 93.22 |
BERTfrozen-BiLSTM-CRF | ✓ | ✗ | ✓ | 93.70 | 94.58 | 94.14 |
BERTfrozen-IDCNN-CRF | ✓ | ✗ | 93.26 | 93.10 | 93.18 | |
AgBERT-BiLSTM-CRF | ✓ | ✗ | ✓ | 94.01 | 94.68 | 94.34 |
AgBERT-IDCNN-CRF | ✓ | ✗ | ✓ | 93.55 | 94.39 | 93.97 |
HNER | ✓ | Subword sequence | ✓ | 93.76 | 93.92 | 93.84 |
