Fig. 2From: Syntax-based transfer learning for the task of biomedical relation extractionThe MCCNN model with three channels, two CNN kernels of size 2 (CNN2) and 3 (CNN3). Red words correspond to the entitiesBack to article page