Table 4 F1 scores with standard deviations on CrossNER.
From: UnionPromptNER serves as a union prompting method to bridge few-shot named entity recognition
Models | CoNLL03 | GUM | WNUT | OntoNotes | Avg. |
|---|---|---|---|---|---|
1-shot | |||||
DecomposedMetaNER | 45.83±0.25 | 16.34±0.33 | 25.01±0.68 | 34.76±0.24 | 30.49 |
SimBERT | 19.22 | 6.91 | 5.18 | 13.99 | 11.32 |
TransferBERT | 4.75±1.42 | 0.57±0.32 | 2.71±0.72 | 3.46±0.54 | 2.87 |
ProtoBERT | 32.49±2.01 | 3.89±0.24 | 10.68±1.40 | 6.67±0.46 | 13.43 |
Matching Network | 19.50±0.35 | 4.73±0.16 | 17.23±2.75 | 15.06±1.61 | 14.13 |
L-TapNET+CDT | 44.30±3.15 | 12.04±0.65 | 20.80±1.06 | 15.17±1.25 | 23.08 |
UnionPromptNER | 46.31±0.61 | 19.58±0.45 | 27.62±0.87 | 35.69±0.53 | 32.3 |
5-shot | |||||
DecomposedMetaNER | 58.04±0.63 | 30.68±0.46 | 31.03±2.04 | 45.61±0.77 | 41.34 |
SimBERT | 32.01 | 10.63 | 8.20 | 21.14 | 18 |
TransferBERT | 15.36±2.81 | 3.62±0.57 | 11.08±0.57 | 35.49±7.60 | 16.39 |
ProtoBERT | 50.06±1.57 | 9.54±0.4 4 | 17.26±2.65 | 13.59±1.61 | 22.61 |
Matching Network | 19.85±0.74 | 5.58±0.23 | 6.61±1.75 | 8.08±0.47 | 10.03 |
L-TapNET+CDT | 45.35±2.67 | 11.65±2.34 | 23.30±2.80 | 20.95±2.81 | 25.32 |
UnionPromptNER | 62.04±0.33 | 42.06±0.88 | 30.54±0.67 | 47.98±0.18 | 45.68 |
