Bridging Discourse Treebanks with a Unified Rhetorical Structure Parser

2025-10-07

AI Summary

Cross-lingual RST treebanks employ incompatible rhetorical relation taxonomies, hindering the development of a unified discourse parser. Method: We propose UniRST—the first end-to-end unified rhetorical structure parsing framework supporting 11 languages and 18 heterogeneous treebanks. Its core innovations include: (1) a Masked-Union training strategy that implicitly aligns disparate relation labels; (2) a multi-head classification mechanism enabling treebank-adaptive prediction under parameter sharing; and (3) data augmentation to enhance generalization for low-resource languages. Results: UniRST outperforms prior methods on 16 of the 18 single-treebank benchmarks. It is the first approach to demonstrate high-performance, multilingual discourse parsing without modifying original relation schemas—thereby validating both the effectiveness and scalability of unified rhetorical parsing across diverse linguistic and annotation paradigms.

Technology Category

• Natural Language Processing: Discourse, Pragmatics & Argument Mining
• Reasoning under Uncertainty: Relational Probabilistic Models
• Knowledge Representation and Reasoning: Argumentation

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

We introduce UniRST, the first unified RST-style discourse parser capable of handling 18 treebanks in 11 languages without modifying their relation inventories. To overcome inventory incompatibilities, we propose and evaluate two training strategies: Multi-Head, which assigns separate relation classification layer per inventory, and Masked-Union, which enables shared parameter training through selective label masking. We first benchmark monotreebank parsing with a simple yet effective augmentation technique for low-resource settings. We then train a unified model and show that (1) the parameter efficient Masked-Union approach is also the strongest, and (2) UniRST outperforms 16 of 18 mono-treebank baselines, demonstrating the advantages of a single-model, multilingual end-to-end discourse parsing across diverse resources.

Problem

• Unified multilingual RST parser handles 18 treebanks
• Resolves inventory incompatibilities across discourse treebanks
• Enables single-model parsing for diverse languages and resources

Innovation

• Unified RST parser handles 18 multilingual treebanks
• Masked-Union training enables shared parameter learning
• Single multilingual model outperforms most mono-treebank baselines