MELD is a multilingual and multi-domain dataset for Named Entity Recognition (NER) constructed from 60 existing datasets.