Fig. 2
From: A sui generis QA approach using RoBERTa for adverse drug event identification
Q&A Module for Adverse Drug Event Identification. Input segment A consists of a drug acting as an alias for question and segment B consists of context where an ADE is mentioned. Encoded representation of segments passed as an input through 12-layered transformer network. 1-D convolutional layer is applied on top of 768 dimensional embedded representations followed by a softmax activation to identify the ADE