cjvt/roberta-en-hi-codemixed

en-hi-codemixed

This is a masked language model, based on the CamemBERT model architecture. en-hi-codemixed model was trained from scratch on English, Hindi, and codemixed English-Hindi corpora for 40 epochs. The corpora used consists of primarily web crawled data, including codemixed tweets, and focuses on conversational language and covid-19 pandemic.