Skip to main content

Table 2 Comparison of performance for comparative methods on the NCBI, GM, and CDR corpora using strict and partial matching and IOB tag matching

From: Biomedical named entity recognition using deep neural networks with contextual information

Modelprfprfprf
Strict matchingNCBIGMCDR
BiLSTM78.9182.6080.7172.2272.4472.3383.5680.2681.88
BiLSTM-CRF82.1984.5883.3780.7979.8180.3087.5283.5885.50
GRAM-CNN84.4583.9284.1880.2378.8379.5386.0885.4985.79
BERT81.0780.7380.9081.7281.5981.6586.2185.2385.72
CLSTMword level85.9484.6985.3181.0080.7780.8987.2385.5186.36
 character level85.4084.0684.7281.0980.3880.7387.1984.6985.92
 word+char levels84.7386.6785.6881.7581.1481.4487.2585.6686.44
Partial matchingNCBIGMCDR
BiLSTM86.6790.7388.6587.9888.2588.1191.1487.5489.30
BiLSTM-CRF91.1993.8592.5193.1892.0492.6194.2790.0092.08
GRAM-CNN94.3693.7894.0793.0991.4792.2792.4791.8392.15
BERT88.3988.0288.2092.6592.5192.5891.8290.7791.29
CLSTMword level93.6692.2992.9792.8192.5492.6793.6091.7492.66
 character level93.7692.2993.0293.0592.2592.6593.4291.5992.49
 word+char levels93.7193.1393.4293.3592.6593.0093.4891.7792.62
IOB tag matchingNCBIGMCDR
BiLSTM84.5688.0386.2684.2381.4882.8389.8178.6883.87
BiLSTM-CRF84.1388.3286.1888.3484.4786.3690.5481.3485.69
GRAM-CNN88.7386.5987.6587.7584.0985.8989.7283.0386.24
BERT88.4283.1585.7089.5086.2687.8588.6985.0186.81
CLSTMword level89.1889.0189.1088.9984.8986.8989.9983.1986.45
 character level88.2188.4788.3487.7285.2186.4589.9583.1986.43
 word+char levels89.9887.7488.8487.1386.8887.0090.5683.4186.83