Skip to main content

Table 8 Comparison results (%accuracy) on discharge summaries

From: A multitask bi-directional RNN model for named entity recognition on Chinese electronic medical records

Model Entity type
  Disease Symptom Disease group Treatment Test Overall accuracy
Naive Bayes (NB) 44.82 51.72 N/A 59.00 65.96 58.91
Maximum Entropy (ME) 48.32 56.34 34.19 58.80 76.10 65.68
Support Vector Machine (SVM) 57.18 62.52 37.22 60.48 80.17 70.46
Conditional Random Field (CRF) [7] 77.33 77.83 48.39 77.47 90.05 83.94
Convolutional Neural Network(CNN) [7] 52.80 65.76 40.00 53.14 79.28 68.60
Bi-RNN model 73.83 79.35 28.00 67.99 82.63 77.85
Transfer learning Bi-RNN model [24] 74.30 82.60 44.00 68.20 86.79 80.75
Our proposed model 76.86 87.22 36.00 71.33 89.20 83.51