Adaptive multi-agent translation improves Bai Chinese in Bowen script
MAS-Bai: Linguistically Informed Adaptive Multi-Agent Translation for Bowen-Script Bai
Computation and Language
Summary
Translating very rare languages like Bai, especially when written in the unusual Bowen script, is hard because existing AI models get confused and make mistakes. The authors created MAS-Bai, a new system that uses multiple language experts and smart routing to handle tricky parts like borrowed Chinese characters and different word orders. They built a special dataset and showed that MAS-Bai translates much better than previous methods. This approach helps ensure translations are more accurate and reliable.
What this means in practice
- •For language technology teams: Develop more reliable machine translation tools for low-resource languages with unique scripts like Bowen-script Bai.
- •For localisation service providers: Produce accurate translations involving Bai and Chinese to support regional content creation and communication.$Commercial implications: Enables translation services for an underserved language market with high-quality results using the MAS-Bai system.
Authors
Jingting Zheng, Hong Ren, Yuehan Cui, Shaolin Zhu, Deyi Xiong
Abstract
Large Language Models (LLMs) remain unreliable for extremely low-resource languages whose writing systems conflict with high-resource pretraining priors. We study Bai-Chinese translation in the Bowen script, where Chinese characters are frequently borrowed phonetically rather than semantically, causing severe literal-translation hallucinations. We propose MAS-Bai, a linguistically informed adaptive multi-agent framework for Bowen-script Bai-Chinese translation. MAS-Bai routes each input through either a lightweight retrieval-translation path or a deeper philological path with etymological and syntactic agents, and uses cycle-consistency with a lightweight reverse translator as an inference-time reliability signal. We construct a 6,536-sentence Bai-Chinese corpus and evaluate on a 1,536-sentence test set. MAS-Bai achieves 16.23 BLEU and 62.15 COMET, improving over Standard RAG by 8.11 BLEU and 6.85 COMET. Forward LoRA, expert ablations and controlled routing experiments distinguish the roles of supervised adaptation, Bai-specific expertise and conditional routing. The largest gains occur on Bowen phonetic loans and divergent Bai word order, while human evaluation and reverse-verifier analyses support the automatic results.