A Minimal Span-Based Neural Constituency Parser
arXiv:1705.03919
Abstract
In this work, we present a minimal neural model for constituency parsing based on independent scoring of labels and spans. We show that this model is not only compatible with classical dynamic programming techniques, but also admits a novel greedy top-down inference algorithm based on recursive partitioning of the input. We demonstrate empirically that both prediction schemes are competitive with recent work, and when combined with basic extensions to the scoring model are capable of achieving state-of-the-art single-model performance on the Penn Treebank (91.79 F1) and strong performance on the French Treebank (82.23 F1).
To appear in ACL 2017
References in corpus (8)
- Adam: A Method for Stochastic Optimization
- Neural Machine Translation by Jointly Learning to Align and Translate
- DyNet: The Dynamic Neural Network Toolkit
- Sequence-to-Sequence Learning as Beam-Search Optimization
- A Unified Tagging Solution: Bidirectional LSTM Recurrent Neural Network with Word Embedding
- Simple and Accurate Dependency Parsing Using Bidirectional LSTM Feature Representations
- Training with Exploration Improves a Greedy Stack-LSTM Parser
- Shift-Reduce Constituent Parsing with Neural Lookahead Features
Cited by in corpus (8)
- Unsupervised Question Answering by Cloze Translation
- Jointly Predicting Predicates and Arguments in Neural Semantic Role Labeling
- Two Local Models for Neural Constituent Parsing
- SpanNER: Named Entity Re-/Recognition as Span Prediction
- A Span Selection Model for Semantic Role Labeling
- Exploiting Syntactic Structure for Better Language Modeling: A Syntactic Distance Approach
- Policy Gradient as a Proxy for Dynamic Oracles in Constituency Parsing
- Semantic Role Labeling for Learner Chinese: the Importance of Syntactic Parsing and L2-L1 Parallel Data