Arc-Standard Spinal Parsing with Stack-LSTMs

TL;DR

This paper introduces a neural transition-based parser for spinal trees using Stack-LSTMs, demonstrating adaptability to various dependency styles and highlighting the LSTMs' ability to induce useful states for constituent prediction.

Contribution

The paper proposes a novel neural parser for spinal trees that leverages Stack-LSTMs, advancing dependency-based constituent parsing methods.

Findings

The parser adapts to different dependency styles.

Dependency style has minimal impact on constituent prediction.

LSTMs effectively induce useful parsing states.

Abstract

We present a neural transition-based parser for spinal trees, a dependency representation of constituent trees. The parser uses Stack-LSTMs that compose constituent nodes with dependency-based derivations. In experiments, we show that this model adapts to different styles of dependency relations, but this choice has little effect for predicting constituent structure, suggesting that LSTMs induce useful states by themselves.