The CLaC Discourse Parser at CoNLL-2016

TL;DR

This paper presents the CLaC discourse parser for the CoNLL-2016 shared task, combining machine learning and deep learning methods to identify explicit and non-explicit discourse relations, achieving an F1-score of 0.2106.

Contribution

It introduces a hybrid approach using both traditional machine learning and deep learning for shallow discourse parsing.

Findings

F1-score of 0.2106 on discourse relation identification

0.3110 F1 for explicit relations

0.1219 F1 for non-explicit relations

Abstract

This paper describes our submission "CLaC" to the CoNLL-2016 shared task on shallow discourse parsing. We used two complementary approaches for the task. A standard machine learning approach for the parsing of explicit relations, and a deep learning approach for non-explicit relations. Overall, our parser achieves an F1-score of 0.2106 on the identification of discourse relations (0.3110 for explicit relations and 0.1219 for non-explicit relations) on the blind CoNLL-2016 test set.