Healthcare predictive analytics aids medical decision-making, diagnosisprediction and drug review analysis . We proposeBBAEG (Biomedical BERT-based Adversarial Example Generation), a black-boxattack algorithm for biomedical text classification . We demonstrate that BBAEGperforms stronger attack with better language fluency, semantic coherence ascompared to prior work . The algorithm leverages the strengthsof both domain-specific synonym replacement for biomedical named entities andBERTMLM predictions, spelling variation and number replacement . It is the latest attempt to generate adversaries usingrule-based synonyms and BERTMLMs .

Author(s) : Ishani Mondal

Links : PDF - Abstract

Code :
Coursera

Keywords : biomedical - based - adversarial - replacement - classification -

Leave a Reply

Your email address will not be published. Required fields are marked *