[4Pin1] インタラクティブ(2)

[4Pin1-30] The Estimation of Argument Structure using Deep Learning

〇Ryo Kinoshita¹, Masaki Uto¹, Maomi Ueno¹ (1. The University of Electro-Communications)

Keywords:Argument Mining, Argument Structure, Natural Language Processing, Deep Learning, Long-Short Term Memory

In this article, we present a novel approach for parsing argumentation structures. We classify argument components using Long-Short Term Memory(LSTM). Then, we use the result of component classification as the feature of the next task, identifying argumentative relations. Finally, the proposed model globally optimizes argumentative relations using integer linear programming whose objective function is the probability estimated in relation identification. Comparative studies show that each of the proposal are effective and our model significantly outperforms the most advanced model.