Generalized chart constraints for efficient PCFG and TAG parsing

TL;DR

This paper introduces generalized chart constraints for more expressive grammar formalisms, utilizing a neural tagger to predict constraints with high precision, significantly speeding up PCFG and TAG parsing while enhancing accuracy.

Contribution

It extends chart constraints to complex grammar formalisms and demonstrates a neural approach for high-precision constraint prediction, leading to substantial parsing speed improvements.

Findings

Achieved two orders of magnitude speedup in parsing

Improved parsing accuracy with combined constraints

Effective integration with other pruning techniques

Abstract

Chart constraints, which specify at which string positions a constituent may begin or end, have been shown to speed up chart parsers for PCFGs. We generalize chart constraints to more expressive grammar formalisms and describe a neural tagger which predicts chart constraints at very high precision. Our constraints accelerate both PCFG and TAG parsing, and combine effectively with other pruning techniques (coarse-to-fine and supertagging) for an overall speedup of two orders of magnitude, while improving accuracy.