Modular neural networks exploit large acoustic context through broad-class posteriors for continuous speech recognition.

Christos Antoniou

Browse the full ICASSP paper archive.