Fast(er) Exact Decoding and Global Training for Transition-Based Dependency Parsing via a Minimal Feature Set
This work addresses efficiency and accuracy in dependency parsing for NLP applications, representing an incremental improvement over prior methods.
The authors tackled the problem of exact decoding and global training for transition-based dependency parsing by introducing a minimal feature set, achieving the best unlabeled attachment score on the Chinese Treebank and a second-best result on the English Penn Treebank.
We first present a minimal feature set for transition-based dependency parsing, continuing a recent trend started by Kiperwasser and Goldberg (2016a) and Cross and Huang (2016a) of using bi-directional LSTM features. We plug our minimal feature set into the dynamic-programming framework of Huang and Sagae (2010) and Kuhlmann et al. (2011) to produce the first implementation of worst-case O(n^3) exact decoders for arc-hybrid and arc-eager transition systems. With our minimal features, we also present O(n^3) global training methods. Finally, using ensembles including our new parsers, we achieve the best unlabeled attachment score reported (to our knowledge) on the Chinese Treebank and the "second-best-in-class" result on the English Penn Treebank.