Global information-aware argument mining based on a top-down multi-turn QA model

Boyang Liu, Viktor Schlegel, Paul Thompson, Riza Theresa Batista-Navarro, Sophia Ananiadou

Abstract: Highlights•Three different types of global information designed for argumentative graph.•A two-step multi-turn QA model explicitly utilising global information to solve the argument mining task.•SOTA on two argument mining datasets and comparable results on the third dataset.