Skip to main content

Table 2 F1 performance of STL, MTL-XC, and MTL-LS with different slicing rates

From: Hierarchical shared transfer learning for biomedical named entity recognition

Dataset STL MTL-XC Slicng rate
0.25 0.50 0.75 1.00
Chemical
BC4CHEMD 91.70 91.25 91.82 92.24 92.37 92.47
BC5CDR 93.36 93.19 93.93 93.75 93.60 93.14
BioNLP11ID 62.94 67.87 42.29 52.98 73.41 73.71
BioNLP13CG 84.16 81.61 78.59 78.50 85.13 82.47
BioNLP13PC 89.44 80.07 81.25 82.13 84.03 86.00
CRAFT 82.54 73.83 76.73 79.74 77.96 79.44
Disease
BC5CDR 86.44 86.87 86.10 87.04 86.28 87.34
NCBI-disease 88.26 88.75 85.23 86.79 89.24 88.97
Gene and protein
BC2GM 82.22 82.64 79.85 81.54 81.94 82.94
BioNLP09 87.50 89.53 87.23 89.53 89.47 89.27
BioNLP11EPI 85.81 86.46 84.65 85.65 86.24 86.58
BioNLP11ID 89.43 89.03 83.62 82.69 89.59 86.55
BioNLP13CG 90.39 92.58 91.33 92.05 92.71 92.21
BioNLP13GE 78.22 83.59 81.29 82.61 82.73 84.03
BioNLP13PC 91.85 89.27 89.16 91.87 90.59 90.37
CRAFT 77.88 81.57 79.02 82.62 83.42 82.18
Ex-PTM 77.52 84.25 78.52 80.57 84.64 84.55
JNLPBA 75.80 78.32 75.08 77.00 77.32 77.66
Species
BioNLP11ID 83.68 79.37 70.91 77.45 83.74 79.00
BioNLP13CG 88.76 86.30 88.97 87.39 86.07 86.90
CRAFT 97.08 93.63 95.56 95.04 94.74 95.75
LINNAEUS 88.56 85.17 83.82 84.05 86.40 85.06
  1. Better scores of each metric are in bold