README.md

Unormalized AMI

from datasets import load_dataset

ami = load_dataset("speech-seq2seq/ami", "ihm")

TODO(PVP) - explain exactly what normalization was accepted what wasn't