Skip to main content

Table 2 Area under learning curve (ALC) for different methods aggregated over all diseases

From: When BERT meets Bilbo: a learning curve analysis of pretrained language model on disease classification

Method

HaoDaiFu (all 805 diseases)

ChinaRe (all 44 diseases)

BOW

0.4158

0.8534

BOW_EXP

0.4266a

0.8934a

BOW_EXP_KG

0.4254a

0.8940a

CBOW

0.2097

0.5817

CBOW_KG

0.2064

0.5714

LSTM

0.2013

0.6064

LSTM_KG

0.0377

0.6243

BERT

0.5020ab

0.9551ab

  1. Figure 2 plots the learning curves
  2. aResult significantly higher than BOW
  3. bResult significantly higher than BOW_EXP_KG. (Fisher's randomization test, significance level \({\upalpha } = 0.05\))