Search papers, labs, and topics across Lattice.
1
0
3
You can shrink a state-of-the-art streaming ASR model by 73% with minimal accuracy loss, opening the door for high-quality, low-latency speech recognition on resource-constrained edge devices.
