LegalRAG: A Hybrid RAG System for Multilingual Legal Information Retrieval

2025-04-19

AI Summary

To address the low retrieval accuracy and poor interpretability in multilingual legal information retrieval from Bangladesh Police Gazette (English–Bengali bilingual documents), this paper proposes a rule-augmented hybrid RAG framework. The method integrates BERT-multilingual for semantic encoding and FAISS for efficient vector-based retrieval, augmented with language identification, bilingual paragraph alignment, and domain-specific keyword enhancement modules. Crucially, it introduces a rule-driven cross-lingual paragraph re-ranking mechanism to improve relevance and transparency. Evaluated on a real-world test set, the approach achieves significant improvements over baseline RAG: Recall@5 increases by 23.6% and answer accuracy rises by 18.4%. It is particularly effective in low-resource bilingual legal settings, enabling real-time, precise, and auditable legal clause localization and question answering.

Technology Category

• Natural Language Processing: Question Answering
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Natural Language Processing (NLP) and computational linguistic techniques are increasingly being applied across various domains, yet their use in legal and regulatory tasks remains limited. To address this gap, we develop an efficient bilingual question-answering framework for regulatory documents, specifically the Bangladesh Police Gazettes, which contain both English and Bangla text. Our approach employs modern Retrieval Augmented Generation (RAG) pipelines to enhance information retrieval and response generation. In addition to conventional RAG pipelines, we propose an advanced RAG-based approach that improves retrieval performance, leading to more precise answers. This system enables efficient searching for specific government legal notices, making legal information more accessible. We evaluate both our proposed and conventional RAG systems on a diverse test set on Bangladesh Police Gazettes, demonstrating that our approach consistently outperforms existing methods across all evaluation metrics.

Problem

• Develops bilingual QA framework for legal documents
• Enhances retrieval performance in multilingual legal texts
• Improves accessibility of government legal notices

Innovation

• Hybrid RAG system for bilingual legal retrieval
• Advanced RAG improves retrieval precision
• Efficient searching for government legal notices