Step-by-step Instructions and a Simple Tabular Output Format Improve the Dependency Parsing Accuracy of LLMs

TL;DR

This paper introduces a step-by-step instruction approach with a simplified output format that significantly improves dependency parsing accuracy of large language models across multiple languages, demonstrating the value of explicit reasoning in structured prediction tasks.

Contribution

The paper presents a novel step-by-step prompting strategy combined with a simplified output format that achieves state-of-the-art dependency parsing accuracy in LLMs across 17 languages.

Findings

Achieves state-of-the-art accuracy on Universal Dependencies datasets

Multilingual fine-tuning enhances cross-language performance

Explicit reasoning steps improve structured prediction in LLMs

Abstract

Recent advances in large language models (LLMs) have enabled impressive performance in various tasks. However, standard prompting often struggles to produce structurally valid and accurate outputs, especially in dependency parsing. We propose a novel step-by-step instruction strategy, where universal part-of-speech tagging precedes the prediction of syntactic heads and dependency labels, and a simplified CoNLL-U like output format, our method achieves state-of-the-art accuracy on Universal Dependencies datasets across 17 languages without hallucination or contamination. We further show that multilingual fine-tuning simultaneously improves cross-language generalization performance. Our results highlight the effectiveness of explicit reasoning steps in LLM-based parsing and offer a scalable, format-consistent alternative to bracket-based approaches.