Fetching the paper…

Have best of both worlds: two-pass hybrid and E2E cascading framework for speech recognition · Around