Skip to main content

Table 4 Cosine similarity score between target and source corpora for the three different pattern distributions

From: Syntax-based transfer learning for the task of biomedical relation extraction

Target corpora

Source corpora

 

DDI

ADE

reACE

 

POS + DT

SNPPhena

0.53

0.22

0.13

EU-ADR

0.24

0.20

0.09

 

POS only

SNPPhena

0.80

0.70

0.35

EU-ADR

0.77

0.68

0.32

 

DT only

SNPPhena

0.53

0.23

0.14

EU-ADR

0.25

0.24

0.10

  1. POS is part of speech pattern and DT is dependency type pattern