Linguistic Hierarchy

Linguistic hierarchy investigates how language structures itself into nested levels, from individual words to complex sentences, reflecting cognitive processes and communicative needs. Current research focuses on computationally inducing these hierarchies from large language models (LLMs), often using transformer architectures and comparing constituency and dependency parsing approaches, to better understand how LLMs represent syntactic and semantic information. This work has implications for improving machine translation, particularly in low-resource languages, and for developing more robust and nuanced natural language processing systems. Furthermore, understanding the hierarchical organization of syntax across time reveals how communicative pressures shape language evolution.