Span-Based Constituency Parsing with a Structure-Label System and Provably Optimal Dynamic Oracles

2016/12/20 by James Cross, Liang Huang, Cross, James +1 · 1 citation
Computer Science · #Computation and Language (cs.CL) #FOS: Computer and information sciences #Natural Language Processing Techniques #Speech Recognition and Synthesis #Topic Modeling

paper · pdf · doi:10.48550/arxiv.1612.06475

openalex publication_date 2016/12/20 · openalex created_date 2019/07/12 · openalex updated_date 2026/07/28

Abstract

Parsing accuracy using efficient greedy transition systems has improved dramatically in recent years thanks to neural networks. Despite striking results in dependency parsing, however, neural models have not surpassed state-of-the-art approaches in constituency parsing. To remedy this, we introduce a new shift-reduce system whose stack contains merely sentence spans, represented by a bare minimum of LSTM features. We also design the first provably optimal dynamic oracle for constituency parsing, which runs in amortized O(1) time, compared to O(n³) oracles for standard dependency parsing. Training with this oracle, we achieve the best F1 scores on both English and French of any parser that does not use reranking or external data.

Citations

• ADADELTA: An Adaptive Learning Rate Method
• Improving neural networks by preventing co-adaptation of feature detectors

Cited by

• In-Order Transition-based Constituent Parsing

Related

• A Minimal Span-Based Neural Constituency Parser
• Incremental Parsing with Minimal Features Using Bi-Directional LSTM
• Straight to the Tree: Constituency Parsing with Neural Syntactic Distance
• Neural Combinatory Constituency Parsing
• Linear-Time Constituency Parsing with RNNs and Dynamic Programming