Learning Syntax from Naturally-Occurring Bracketings

TL;DR

This paper introduces a novel approach to unsupervised constituency parsing by leveraging naturally-occurring bracketings like answer fragments and hyperlinks, using a partial-brackets-aware structured ramp loss to improve accuracy.

Contribution

It proposes a new method that effectively incorporates noisy and incomplete natural bracketings into unsupervised parsing models, achieving state-of-the-art results.

Findings

Achieved an unlabeled F1 score of 68.9 on the WSJ corpus.

Models outperform existing unsupervised systems in syntactic structure induction.

Demonstrated the utility of naturally-occurring bracketings as distant supervision signals.

Abstract

Naturally-occurring bracketings, such as answer fragments to natural language questions and hyperlinks on webpages, can reflect human syntactic intuition regarding phrasal boundaries. Their availability and approximate correspondence to syntax make them appealing as distant information sources to incorporate into unsupervised constituency parsing. But they are noisy and incomplete; to address this challenge, we develop a partial-brackets-aware structured ramp loss in learning. Experiments demonstrate that our distantly-supervised models trained on naturally-occurring bracketing data are more accurate in inducing syntactic structures than competing unsupervised systems. On the English WSJ corpus, our models achieve an unlabeled F1 score of 68.9 for constituency parsing.