Holistic Exploration on Universal Decompositional Semantic Parsing: Architecture, Data Augmentation, and LLM Paradigm

A long-standing objective in the fields of natural Secondly, while Stengel-Eskin et al. language understanding and computational semantics (2020) have tried to utilize the external tool Pred-is to create a structured graph of linguistic Patt (Zhang et al., 2017), which contains the relationship meaning. Various efforts have been made to between syntax and semantic information, encode semantic relations and attributes into a semantic they do not achieve any improvements. In graph--e.g., Abstract Meaning Representation contrast, we propose a data augmentation method (AMR; Banarescu et al., 2013), Universal that effectively exploits the capabilities of Pred-Conceptual Cognitive Annotation (UCCA; Abend Patt, leading to significant performance gains in and Rappoport, 2013), and Semantic Dependency relation parsing. Moreover, we have explored various Parsing formalisms (SDP; Oepen et al., 2014, approaches for these enhancements, providing 2016).