Transmi: A Machine Learning Framework To Create Baseline Models Adapted For Transliterated Data From Existing Multilingual Pretrained Language Models Mplms Without Any Training

TRANSMI is a framework designed to enhance multilingual pre-trained language models (mPLMs) for transliterated data, addressing issues with cross-lingual transfer learning and improving NLP applications across various languages and scripts. It modifies existing mPLMs using three merge modes to incorporate transliterated subwords into their vocabularies, resulting in improved performance in cross-lingual tasks.