Sentence-Level Discourse Parsing as Text-to-Text Generation

Anonymous

Abstract: Previous studies have made great advances in RST discourse parsing through neural frameworks or efficient features, but they split the parsing process into two subtasks and heavily depended on gold segmentation. In this paper, we introduce an end-to-end method for sentence-level RST discourse parsing via transforming it into a text-to-text generation task. Our method unifies the traditional two-stage parsing and generates the parsing tree directly from the input text without requiring a complicated model. Moreover, the EDU segmentation can be simultaneously generated and extracted from the parsing tree. Experimental results on the RST Discourse Treebank demonstrate that our proposed method outperforms existing methods in both tasks of sentence-level RST parsing and discourse segmentation. Considering the lack of annotated data in RST parsing, we also create high-quality augmented data based on several filtering strategies, which further improves the performance.