Transcending Language Boundaries: Harnessing LLMs for Low-Resource Language Translation

• DOI: 10.48550/arxiv.2411.11295
• Publication Date: 2024-11-18
• Authors: P. Shu; Junhao Chen; Zhengliang Liu; Hui Wang; Zihao Wu; Tianyang Zhong; Yiwei Li; Huaqin Zhao; Hanqi Jiang; Yi Pan; Yifan Zhou; Constance Owl; Xiaoming Zhai; Ninghao Liu; Claudio Saunt; Tianming Liu
• Keywords: FOS: Computer and information sciences; Computer Science - Computation and Language; Artificial Intelligence (cs.AI); Computer Science - Artificial Intelligence; Computation and Language (cs.CL)

Large Language Models (LLMs) have demonstrated remarkable success across a wide range of tasks and domains. However, their performance in low-resource language translation, particularly when translating into these languages, remains underexplored. This gap poses significant challenges, as linguistic barriers hinder the cultural preservation development minority communities. To address this issue, paper introduces novel retrieval-based method that enhances translation quality for languages by focusing on key terms, which involves keywords retrieving corresponding examples from existing data. evaluate effectiveness method, we conducted experiments English three languages: Cherokee, critically endangered indigenous North America; Tibetan, historically culturally Asia; Manchu, with few remaining speakers. Our comparison zero-shot GPT-4o LLaMA 3.1 405B, highlights challenges models face languages. In contrast, our shows promise improving both word-level accuracy overall semantic understanding leveraging resources more effectively.