Cross-topic Argument Mining from Heterogeneous Sources Using Attention-based Neural Networks

TL;DR

This paper introduces an attention-based neural network for argument mining that effectively generalizes across diverse topics and text types, outperforming traditional models in accuracy and F-score.

Contribution

It presents a novel sentential annotation scheme and demonstrates the effectiveness of an attention-based neural network for cross-topic argument mining on heterogeneous web texts.

Findings

Attention-based neural network outperforms BiLSTM by 6% in accuracy.

Model achieves 11% higher F-score on unseen topics.

Crowd-sourced annotations enable scalable argument mining across diverse texts.

Abstract

Argument mining is a core technology for automating argument search in large document collections. Despite its usefulness for this task, most current approaches to argument mining are designed for use only with specific text types and fall short when applied to heterogeneous texts. In this paper, we propose a new sentential annotation scheme that is reliably applicable by crowd workers to arbitrary Web texts. We source annotations for over 25,000 instances covering eight controversial topics. The results of cross-topic experiments show that our attention-based neural network generalizes best to unseen topics and outperforms vanilla BiLSTM models by 6% in accuracy and 11% in F-score.