Thanmay Jayakumar, Deepon Halder, Raj Dabre
View original ↗Develop a unified pre-processing pipeline for transliteration that helps LLMs handle low-resource scripts better. It could serve as a drop-in middleware for existing RAG systems.
Suggested repo: scriptBridge
"Break the script barrier: standardize multilingual inputs before they hit your model."
Estimated effort: 30h