Using Convolutional Neural Network in Cross-Domain Argumentation Mining Framework

Abstract

Argument Mining has become a remarkable research area in computational argumentation and Natural Language Processing fields. Despite its importance, most of the current proposals are restricted to a text type (e.g., Essays, web comments) on a specific domain and fall behind expectations when applied to cross-domain data. This paper presents a new framework for Argumentation Mining to detect argumentative segments and their components automatically using Convolutional Neural Network (CNN). We focus on both (1) argumentative sentence detection and (2) argument components detection tasks. Based on different corpora, we investigate the performance of CNN on both in-domain level and cross-domain level. The investigation shows challenging results in comparison with classic machine learning models.