Joint POS Tagging and Dependency Parsing with Transition-based Neural Networks

摘要

While part-of-speech (POS) tagging and dependency parsing are observed to beclosely related, existing work on joint modeling with manually crafted featuretemplates suffers from the feature sparsity and incompleteness problems. Inthis paper, we propose an approach to joint POS tagging and dependency parsingusing transition-based neural networks. Three neural network based classifiersare designed to resolve shift/reduce, tagging, and labeling conflicts.Experiments show that our approach significantly outperforms previous methodsfor joint POS tagging and dependency parsing across a variety of naturallanguages.