Enhancing Legal Document Retrieval: A Multi-Phase Approach with Large Language Models

TL;DR

This paper proposes a multi-phase legal document retrieval system that combines BM25, BERT re-ranking, and large language model prompting, significantly improving accuracy on the COLIEE 2023 dataset.

Contribution

It introduces a novel multi-phase retrieval framework that effectively integrates prompting with traditional retrieval methods for legal documents.

Findings

Enhanced retrieval accuracy with LLM prompting

Effective combination of BM25, BERT, and LLMs

Identified challenges in current retrieval systems

Abstract

Large language models with billions of parameters, such as GPT-3.5, GPT-4, and LLaMA, are increasingly prevalent. Numerous studies have explored effective prompting techniques to harness the power of these LLMs for various research problems. Retrieval, specifically in the legal data domain, poses a challenging task for the direct application of Prompting techniques due to the large number and substantial length of legal articles. This research focuses on maximizing the potential of prompting by placing it as the final phase of the retrieval system, preceded by the support of two phases: BM25 Pre-ranking and BERT-based Re-ranking. Experiments on the COLIEE 2023 dataset demonstrate that integrating prompting techniques on LLMs into the retrieval system significantly improves retrieval accuracy. However, error analysis reveals several existing issues in the retrieval system that still need resolution.