Multitask Pointer Network for Korean Dependency Parsing

Sangkeun Jung, Cheon-Eum Park, Changki Lee

Abstract

Dependency parsing is a fundamental problem in natural language processing. We introduce a novel dependency parsing framework called head pointing based dependency parsing. In this framework, we cast a Korean dependency parsing to a statistical head pointing and arc labeling problem. To address this problem, a novel neural network called Multitask Pointer Network is proposed for a neural sequential head pointing and type labeling. Our approach does not require any hand-crafting features or language-specific rules to parse the dependency. Furthermore, it demonstrates a state-of-the-art performance in Korean dependency parsing.