N EURAL F ACTORS , AND A PPROXIMATION - AWARE T RAINING by Matthew

This thesis broadens the space of rich yet practical models for structured prediction. Weintroduce a general framework for modeling with four ingredients: (1) latent variables,(2) structural constraints, (3) learned (neural) feature representations of the inputs, and(4) training that takes the approximations made during inference into account. The thesisbuilds up to this framework through an empirical study of three NLP tasks: semantic rolelabeling, relation extraction, and dependency parsing—obtaining state-of-the-art results onthe former two. We apply the resulting graphical models with structured and neural fac-tors, and approximation-aware learning to jointly model part-of-speech tags, a syntacticdependency parse, and semantic roles in a low-resource setting where the syntax is unob-served. We present an alternative view of these models as neural networks with a topologyinspired by inference on graphical models that encode our intuitions about the data.