Table 3.

Summary of deep learning models used

Model code	Model type
PCNN	Piecewise Convolutional Neural Network (PCNN) with pretrained word2vec word vectors
BIOBERT	BioBERT uncased base model with pretrained weights with entity markers
BAG_PCNN	Multi-instance learning PCNN using pretrained word2vec word vectors and with instances having the same entity pairs grouped in bags.
BAG_BIOBERT	Multi-instance learning BioBERT uncased base model with pretrained weights and with instances having the same entity pairs grouped in bags.