Skip to main content

Table 1 Evaluation results on the official test set

From: Deep learning with sentence embeddings pre-trained on biomedical corpora improves the performance of finding similar sentences in electronic medical records

 

# input models

Test set correlation

Original submissions

 Random Forest (submission #1)

1

0.8106

 Random Forest + Dense Network (submission #2)

2

0.8246

 Ensemble model (submission #3)

8

0.8328

 Random Forest + Encoder Network (submission #4)

2

0.8258

Improved models

 Random Forest

1

0.8246

 Encoder Network

1

0.8384

 Ensemble model

3

0.8528