Skip to main content
. 2021 Sep;9(18):1415. doi: 10.21037/atm-21-3828

Table 5. Performance comparison of entity recognition results under different systems.

Category BIO BIEO
BiLSTM + CRF IDCNN + CRF BiLSTM + CRF IDCNN + CRF
P (%) R (%) F1 P (%) R (%) F1 P (%) R (%) F1 P (%) R (%) F1
All 92.03 91.94 0.9201 91.62 91.52 0.9183 90.76 91.40 0.9108 90.92 91.75 0.9178
Disease 86.52 85.87 0.8671 86.76 85.95 0.8701 83.76 84.09 0.8491 83.92 85.12 0.8508
Symptoms 94.06 93.83 0.9425 94.26 93.97 0.9375 93.36 93.33 0.9335 93.86 92.53 0.9401
Location 96.06 96.88 0.9577 96.26 96.68 0.9527 94.26 94.68 0.9447 95.16 95.08 0.9497
Exam 88.78 90.04 0.8975 89.21 89.84 0.8885 88.27 89.04 0.8865 88.36 89.46 0.8950
Exam results 85.25 84.95 0.8397 84.19 84.07 0.8382 82.29 82.51 0.8240 83.01 82.94 0.8259

BIO, Begin-Intermediate-Other sequence annotation; BIEO, Begin-Intermediate-End-Other sequence annotation; BiLSTM, bidirectional long- and short-term memory; CRF, conditional random field; IDCNN, iterated dilated convolutional neural network.