Hierarchical Clause Annotation: Building a Clause-Level Corpus for Semantic Parsing with Complex Sentences

Most natural-language-processing (NLP) tasks suffer performance degradation when encountering long complex sentences, such as semantic parsing, syntactic machine translation, and text summarization. Previous works addressed the issue with intuition of decomposing sentences linking simple ones, rhetorical-structure-theory (RST)-style discourse split-and-rephrase (SPRP), simplification (TS), sentence decomposition (SSD), etc. However, these are not applicable for parsing abstract meaning representation (AMR) dependency due to misalignments relations unavailabilities preserve original semantics. Following same avoiding deficiencies previous works, we propose a novel framework, hierarchical clause annotation (HCA), capturing clausal structures based on linguistic research hierarchy. With HCA annotated large corpus explore potentialities integrating structural features into sentences. Moreover, decomposed two subtasks, i.e., segmentation provide neural baseline models more-silver annotations. In evaluating proposed our manually dataset, performances resulted in 91.3% F1-scores 88.5% Parseval scores, respectively. Due model architectures employed, differences clause/discourse subtasks was reflected compared corpora, where contained more segment units fewer interrelations than those corpora.