AI4Bharat's IndicBERT finetuned for few-shot transfer learning by fine-tuning on Hindi training data with Urdu validation and test sets. Expected low accuracy. Leverages mbert's tokenizer in this implementation.
language:
- ur tags:
- named entity recognition
- ner license: apache-2.0 datasets:
- wikiann metrics:
- f1
- recall
- precision
- accuracy