The new system is based on a sequence-to-sequence network that translates speech using source spectrograms as input.
17 May 2019 11:45 AM IST