Fast Rule-Based Decoding: Revisiting Syntactic Rules in Neural Constituency Parsing

TL;DR

This paper introduces a GPU-accelerated, rule-based CKY decoding method for neural constituency parsing, significantly improving accuracy and cross-domain performance by integrating syntactic rules into neural models.

Contribution

It presents a fast, rule-constrained CKY decoding approach that incorporates syntactic rules into neural constituency parsing, enhancing accuracy and efficiency.

Findings

Achieved 95.89 F1 on PTB dataset

Achieved 92.52 F1 on CTB dataset

Demonstrated strong zero-shot cross-domain performance

Abstract

Most recent studies on neural constituency parsing focus on encoder structures, while few developments are devoted to decoders. Previous research has demonstrated that probabilistic statistical methods based on syntactic rules are particularly effective in constituency parsing, whereas syntactic rules are not used during the training of neural models in prior work probably due to their enormous computation requirements. In this paper, we first implement a fast CKY decoding procedure harnessing GPU acceleration, based on which we further derive a syntactic rule-based (rule-constrained) CKY decoding. In the experiments, our method obtains 95.89 and 92.52 F1 on the datasets of PTB and CTB respectively, which shows significant improvements compared with previous approaches. Besides, our parser achieves strong and competitive cross-domain performance in zero-shot settings.