Kalpit Dixit

UDA for Sequence Classification problems (like NER)

2

Hi, Regarding UDA for NLP-SequenceTagging like NER. Any ideas for what kind of augmentation methods would work for them? e.g. BackTranslation is tough to use for NER because, 1. we...

SentencePiece Tokenizers uses Train file as Test file when using json annotation_format

3

The SentencePiece tokenizer is trained on a train file `annotation_train`. Then the [SentencePiece tokenizer script at L339](https://github.com/speechbrain/speechbrain/blob/develop/speechbrain/tokenizers/SentencePiece.py#L339) uses `annotation_list_to_check list` as a list of input files to check the coverage...