Argument Mining with Structured SVMs and RNNs

TL;DR

This paper introduces a novel factor graph model for argument mining that handles non-tree structures, jointly classifies units and relations, supports SVM and RNN parametrizations, and outperforms baseline methods.

Contribution

It presents a new structured model for argument mining capable of handling complex, non-tree argument structures and integrating different learning paradigms.

Findings

Outperforms unstructured baselines on web comments data

Supports structure constraints like transitivity

Handles non-tree argumentative relations

Abstract

We propose a novel factor graph model for argument mining, designed for settings in which the argumentative relations in a document do not necessarily form a tree structure. (This is the case in over 20% of the web comments dataset we release.) Our model jointly learns elementary unit type classification and argumentative relation prediction. Moreover, our model supports SVM and RNN parametrizations, can enforce structure constraints (e.g., transitivity), and can express dependencies between adjacent relations and propositions. Our approaches outperform unstructured baselines in both web comments and argumentative essay datasets.