A novel ensemble model with two-stage learning for joint dialog act recognition and sentiment classification

Yujun Xu, Enguang Yao, Chaoyue Liu, Qidong Liu, Mingliang Xu

Abstract: Highlights•Proposed a two-stage contextual learning (TSCL) framework for joint DAR and SC.•Improved the classifier’s performance by dynamically reducing the negative effects of unreliable contextual information.•Introduced an edge-aware graph attention network (EAGAT).•It can selectively leverage the contextual information by dynamically generating attention coefficients based on edge features.