GRAFT: A Graph-based Flow-aware Agentic Framework for Document-level Machine Translation

TL;DR

GRAFT introduces a graph-based, discourse-aware framework leveraging LLM agents for document-level machine translation, significantly improving translation quality and coherence across multiple domains and languages.

Contribution

It presents a novel graph-augmented, agentic framework that effectively models discourse dependencies and enhances document translation accuracy.

Findings

Achieves 2.8 BLEU improvement on IWSLT2017 TED test sets.

Attains 2.3 BLEU gain for English-Chinese domain-specific translation.

Demonstrates consistent discourse-level translation improvements.

Abstract

Document level Machine Translation (DocMT) approaches often struggle with effectively capturing discourse level phenomena. Existing approaches rely on heuristic rules to segment documents into discourse units, which rarely align with the true discourse structure required for accurate translation. Otherwise, they fail to maintain consistency throughout the document during translation. To address these challenges, we propose Graph Augmented Agentic Framework for Document Level Translation (GRAFT), a novel graph based DocMT system that leverages Large Language Model (LLM) agents for document translation. Our approach integrates segmentation, directed acyclic graph (DAG) based dependency modelling, and discourse aware translation into a cohesive framework. Experiments conducted across eight translation directions and six diverse domains demonstrate that GRAFT achieves significant performance gains over state of the art DocMT systems. Specifically, GRAFT delivers an average improvement of 2.8 d BLEU on the TED test sets from IWSLT2017 over strong baselines and 2.3 d BLEU for domain specific translation from English to Chinese. Moreover, our analyses highlight the consistent ability of GRAFT to address discourse level phenomena, yielding coherent and contextually accurate translations.