Achieving Human Parity in Conversational Speech Recognition
share
Conversational speech recognition has served as a flagship speech recognition task since the release of the Switchboard corpus in the 1990s. In this paper, we measure the human error rate on the widely used NIST 2000 test set, and find that our latest automated system has reached human parity. The error rate of professional transcribers is 5.9 which newly acquainted pairs of people discuss an assigned topic, and 11.3 the CallHome portion where friends and family members have open-ended conversations. In both cases, our automated system establishes a new state of the art, and edges past the human benchmark, achieving error rates of 5.8 11.0 convolutional and LSTM acoustic model architectures, combined with a novel spatial smoothing method and lattice-free MMI acoustic training, multiple recurrent neural network language modeling approaches, and a systematic use of system combination.
READ FULL TEXT