Abstract
This study introduces an approach for evaluating the importance of signals proposed by Das and Taboada in discourse parsing. Previous studies using other signals indicate that discourse markers (DMs) are not consistently reliable cues and can act as distractors, complicating relations recognition. The study explores the effectiveness of alternative signal types, such as syntactic and genre-related signals, revealing their efficacy even when not predominant for specific relations. An experiment incorporating RST signals as features for a parser error / success prediction model demonstrates their relevance and provides insights into signal combinations that prevents (or facilitates) accurate relation recognition. The observations also identify challenges and potential confusion posed by specific signals. This study resulted in producing publicly available code and data, contributing to an accessible resources for research on RST signals in discourse parsing.- Anthology ID:
- 2024.codi-1.13
- Volume:
- Proceedings of the 5th Workshop on Computational Approaches to Discourse (CODI 2024)
- Month:
- March
- Year:
- 2024
- Address:
- St. Julians, Malta
- Editors:
- Michael Strube, Chloe Braud, Christian Hardmeier, Junyi Jessy Li, Sharid Loaiciga, Amir Zeldes, Chuyuan Li
- Venues:
- CODI | WS
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 139–148
- Language:
- URL:
- https://aclanthology.org/2024.codi-1.13
- DOI:
- Cite (ACL):
- Martial Pastor and Nelleke Oostdijk. 2024. Signals as Features: Predicting Error/Success in Rhetorical Structure Parsing. In Proceedings of the 5th Workshop on Computational Approaches to Discourse (CODI 2024), pages 139–148, St. Julians, Malta. Association for Computational Linguistics.
- Cite (Informal):
- Signals as Features: Predicting Error/Success in Rhetorical Structure Parsing (Pastor & Oostdijk, CODI-WS 2024)
- PDF:
- https://preview.aclanthology.org/nschneid-patch-1/2024.codi-1.13.pdf