Skip to main content

Table 2 Pearson and Spearman correlations coefficient score (\(\times 100\)) between model predictions and human ratings on three evaluation datasets

From: Refining electronic medical records representation in manifold subspace

Method MayoSRS UMNSRS-sim UMNSRS-rel
Pearson Spearman Pearson Spearman Pearson Spearman
BERT 24.7 24.5 28.3 26.2 31.4 28.2
Zhang 62.5 61.1 64.9 62.5 57.0 57.0
Chiu 60.4 61.5 66.3 65.2 60.0 60.1
ALBERT 24.9 25.0 28.7 26.6 31.5 28.7
BioBERT 26.0 25.5 29.8 27.4 33.4 29.4
BlueBERT 26.5 27.6 31.2 28.9 33.9 30.4
Ours 63.2 62.1 67.0 66.5 61.3 60.8
  1. Bold values denote the best result for each column of data