A research project that investigates the linguistic biases introduced by English-centric training data on multilingual language models (LMs). The widespread use of English data for training these models can lead to performance skews and biases when applied to other languages.
fiifidawson/Multilingual-LM-Study
A research project that investigates the linguistic biases introduced by English-centric training data on multilingual language models (LMs). The widespread use of English data for training these models can lead to performance skews and biases when applied to other languages.
Jupyter Notebook