Hierarchical Attention-enhanced Contextual CapsuleNet for Multilingual Hope Speech Detection

Mohammad Zia Ur Rehman, Devraj Raghuvanshi, Harshit Pachar, Chandravardhan Singh Raghaw, Nagendra Kumar

Abstract: Highlights•A novel framework for hope speech detection in low-resource code-mixed languages.•A custom CasuleNet integrates two-level probability vectors via a skip connection.•Capsule-level and hierarchical attention help focus on discriminative information.•Translated, transliterated Dravidian, and Roman versions are taken into account.•Qualitative and quantitative experiments on three low-resource languages.