Transition-Based Syntactic Linearization with Lookahead Features

It has been shown that transition-based methods can be used for syntactic word ordering and tree linearization, achieving significantly faster speed compared with traditional best-first methods. State-of-the-art transitionbased models give competitive results on abstract word ordering and unlabeled tree linearization, but significantly worse results on labeled tree linearization. We demonstrate that the main cause for the performance bottleneck is the sparsity of SHIFT transition actions rather than heavy pruning. To address this issue, we propose a modification to the standard transition-based feature structure, which reduces feature sparsity and allows lookahead features at a small cost to decoding efficiency. Our model gives the best reported accuracies on all benchmarks, yet still being over 30 times faster compared with best-first-search.

[1]  William J. Byrne,et al.  Fluency Constraints for Minimum Bayes-Risk Decoding of Statistical Machine Translation Lattices , 2010, COLING.

[2]  Anja Belz,et al.  The First Surface Realisation Shared Task: Overview and Evaluation Results , 2011, ENLG.

[3]  Stephen Clark,et al.  Syntax-Based Grammaticality Improvement using CCG and Guided Search , 2011, EMNLP.

[4]  Yue Zhang,et al.  An Empirical Comparison Between N-gram and Syntactic Language Models for Word Ordering , 2015, EMNLP.

[5]  Danqi Chen,et al.  A Fast and Accurate Dependency Parser using Neural Networks , 2014, EMNLP.

[6]  Leo Wanner,et al.  Broad Coverage Multilingual Deep Sentence Generation with a Stochastic Multi-Level Realizer , 2010, COLING.

[7]  Joakim Nivre,et al.  Deterministic Dependency Parsing of English Text , 2004, COLING.

[8]  Stephen Wan,et al.  Improving Grammaticality in Statistical Sentence Generation: Introducing a Dependency Spanning Tree Algorithm with an Argument Satisfaction Model , 2009, EACL.

[9]  Yue Zhang,et al.  Transition-Based Syntactic Linearization , 2015, NAACL.

[10]  Stephen Clark,et al.  Syntactic Processing Using the Generalized Perceptron and Beam Search , 2011, CL.

[11]  Yue Zhang,et al.  Joint Morphological Generation and Syntactic Linearization , 2014, AAAI.

[12]  Marcus Tomalin,et al.  Word Ordering with Phrase-Based Grammars , 2014, EACL.

[13]  Stephen Clark,et al.  Discriminative Syntax-Based Word Ordering for Text Generation , 2015, CL.

[14]  Stephen Clark,et al.  Syntax-Based Word Ordering Incorporating a Large-Scale Language Model , 2012, EACL.

[15]  Michael Strube,et al.  Generating Constituent Order in German Clauses , 2007, ACL.

[16]  Michael White,et al.  Reining in CCG Chart Realization , 2004, INLG.

[17]  Wei He,et al.  Dependency Based Chinese Sentence Realization , 2009, ACL/IJCNLP.

[18]  Michael White,et al.  Perceptron Reranking for CCG Realization , 2009, EMNLP.

[19]  Yue Zhang.  Partial-Tree Linearization: Generalized Word Ordering for Text Synthesis , 2013, IJCAI.

[20]  Yue Zhang,et al.  Syntactic SMT Using a Discriminative Text Generation Model , 2014, EMNLP.