Fig. 1
From: A sui generis QA approach using RoBERTa for adverse drug event identification
End-to-end system architecture. (a) Positive (ADE) and Negative (Non-ADE) sentences are passed through NER module (b) NER module filters sentences which don’t have drug entity mentions (c) Selected sentences at NER stage are passed through Classification Module which filters out sentences where Drug-ADE relation is less probable (d) RoBERTa QA Module identifies ADE relationship (e.g. depressed level of consciousness) corresponding to the selected drug entity (e.g. ibuprofen)