A segmental framework for fully-unsupervised large-vocabulary speech recognition | Read Paper on Bytez