Solomon at SemEval-2020 Task 11: Ensemble Architecture for Fine-Tuned Propaganda Detection in News Articles

This paper describes our system (Solomon) details and results of\nparticipation in the SemEval 2020 Task 11 "Detection of Propaganda Techniques\nin News Articles"\\cite{DaSanMartinoSemeval20task11}. We participated in Task\n"Technique Classification" (TC) which is a multi-class classification task. To\naddress the TC task, we used RoBERTa based transformer architecture for\nfine-tuning on the propaganda dataset. The predictions of RoBERTa were further\nfine-tuned by class-dependent-minority-class classifiers. A special classifier,\nwhich employs dynamically adapted Least Common Sub-sequence algorithm, is used\nto adapt to the intricacies of repetition class. Compared to the other\nparticipating systems, our submission is ranked 4th on the leaderboard.\n