1 paper · 1 filter
Yu Nakagome, Michael Hentschel
Despite recent advances in end-to-end speech recognition methods, their output is biased to the training data's vocabulary, resulting in inaccurate recognition of unknown terms or…