Fast(er) Exact Decoding and Global Training for Transition-Based Dependency Parsing via a Minimal Feature Set
Fast(er) Exact Decoding and Global Training for Transition-Based Dependency Parsing via a Minimal Feature Set
复制标题
DOI：
10.18653/v1/d17-1002
复制
发表时间：
2017-08
期刊：
影响因子：
--
通讯作者：
Tianze Shi;Liang Huang;Lillian Lee
中科院分区：
文献类型：
--
作者：
Tianze Shi;Liang Huang;Lillian Lee
我们首先提出了一个用于基于转换的依赖解析的最小特征集，延续了Kiperwasser和Goldberg（2016 a）以及Cross和Huang（2016 a）发起的使用双向LSTM特征的最新趋势。我们将我们的最小特征集插入Huang和Sagae（2010）和Kuhlmann et al.（2011）的动态编程框架中，以产生弧混合和弧渴望转换系统的最坏情况O（n^3）精确解码器的第一个实现。利用我们的最小特征，我们还提出了O（n^3）的全局训练方法。最后，使用包括我们的新解析器在内的集合，我们在中文树库中获得了（据我们所知）报告的最佳未标记附件分数，在英文宾夕法尼亚树库中获得了“同类第二佳”结果。
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.