Recent Advances in Transition-Based Dependency Parsing - 2012

Explore recent advancements in transition-based dependency parsing in this comprehensive lecture by Professor Joakim Nivre. Delve into the evolution of systems like MaltParser, which achieve linear-time parsing with projective dependency trees using locally trained classifiers and greedy best-first search. Discover how globally trained classifiers and beam search can mitigate error propagation and enable richer feature representations. Examine various methods for extending coverage to non-projective trees, crucial for linguistic adequacy in many languages. Learn about a novel model for joint tagging and parsing that improves both tagging and parsing accuracy compared to standard pipeline approaches. Gain insights from Nivre's extensive research in computational linguistics, including his work on the transition-based approach to syntactic dependency parsing and the development of the MaltParser system.