Towards Unification of Discourse Annotation Frameworks

TL;DR

This paper explores unifying major discourse annotation frameworks (RST, PDTB, SDRT) to enable better use of existing corpora and improve downstream NLP tasks through systematic relations and automatic unification methods.

Contribution

It proposes a comprehensive approach to unify discourse structure and relations across frameworks, including evaluation and applications in multi-task learning and graphical models.

Findings

Preliminary analysis of framework relations

Proposed automatic unification methods

Potential improvements in downstream tasks

Abstract

Discourse information is difficult to represent and annotate. Among the major frameworks for annotating discourse information, RST, PDTB and SDRT are widely discussed and used, each having its own theoretical foundation and focus. Corpora annotated under different frameworks vary considerably. To make better use of the existing discourse corpora and achieve the possible synergy of different frameworks, it is worthwhile to investigate the systematic relations between different frameworks and devise methods of unifying the frameworks. Although the issue of framework unification has been a topic of discussion for a long time, there is currently no comprehensive approach which considers unifying both discourse structure and discourse relations and evaluates the unified framework intrinsically and extrinsically. We plan to use automatic means for the unification task and evaluate the result with structural complexity and downstream tasks. We will also explore the application of the unified framework in multi-task learning and graphical models.