Structure-Preserving Document Translation via Multi-Stage LLM Pipeline: A Case Study in Marathi

2026-06-27

AI Summary

This study addresses structural degradation, formatting inconsistencies, and terminological discrepancies commonly encountered in translating official Marathi documents from Maharashtra, India, into English. To this end, the authors propose an end-to-end, high-fidelity document translation framework that integrates layout-aware OCR, coordinate-driven text extraction, large language model (LLM)-based translation, and HTML-based structural reconstruction. Notably, the approach introduces spatial alignment constraints and hierarchical structure preservation mechanisms within a multi-stage LLM translation pipeline—a novel contribution to the field. Experimental results on real-world government PDF documents demonstrate that the proposed method substantially outperforms conventional plain-text translation in preserving both visual and logical document structure, enhancing translation coherence, and improving domain-specific terminology accuracy.

Abstract

Government documents in India are predominantly issued in regional languages such as Marathi, creating substantial accessibility barriers for non-native readers, interstate administrative bodies, and policy analysts. Although recent advances in neural machine translation have improved sentence-level translation quality, existing systems largely neglect document structure, formatting integrity, and domain-specific terminology, thereby limiting their applicability to official documentation. This paper presents a structure-preserving Marathi-to-English government document translation framework capable of performing end-to-end document transformation while maintaining layout fidelity. The proposed system integrates layout-aware optical character recognition, coordinate-based text extraction, large language model based translation, and structured document reconstruction through HTML representations. By enforcing spatial alignment constraints and preserving hierarchical document elements, the framework ensures structural consistency between the source and translated documents. Experimental evaluation on real-world Marathi government PDFs demonstrates improved structural preservation, translation coherence, and terminological consistency compared to conventional text-only translation pipelines. The proposed framework contributes toward scalable multilingual accessibility solutions for e-governance and administrative document processing.

Problem

• document translation
• structure preservation
• government documents
• Marathi
• accessibility

Innovation

• structure-preserving translation
• multi-stage LLM pipeline
• layout-aware OCR
• document reconstruction
• terminological consistency