Second-Order Unsupervised Neural Dependency Parsing

TL;DR

This paper introduces a second-order unsupervised neural dependency parsing model that incorporates grandparent and sibling information, improving parsing accuracy over first-order models and state-of-the-art benchmarks.

Contribution

It proposes a novel second-order extension with a new neural parameterization and a joint training framework to handle lexicalization efficiently.

Findings

Second-order models outperform first-order in unsupervised parsing.

Joint training improves lexicalized and unlexicalized model performance.

Achieves 10% higher accuracy on WSJ dataset than previous best.

Abstract

Most of the unsupervised dependency parsers are based on first-order probabilistic generative models that only consider local parent-child information. Inspired by second-order supervised dependency parsing, we proposed a second-order extension of unsupervised neural dependency models that incorporate grandparent-child or sibling information. We also propose a novel design of the neural parameterization and optimization methods of the dependency models. In second-order models, the number of grammar rules grows cubically with the increase of vocabulary size, making it difficult to train lexicalized models that may contain thousands of words. To circumvent this problem while still benefiting from both second-order parsing and lexicalization, we use the agreement-based learning framework to jointly train a second-order unlexicalized model and a first-order lexicalized model. Experiments on multiple datasets show the effectiveness of our second-order models compared with recent state-of-the-art methods. Our joint model achieves a 10% improvement over the previous state-of-the-art parser on the full WSJ test set