Search papers, labs, and topics across Lattice.
AGIBOT
1
0
3
Margin loss fine-tuning of ECAPA-TDNNs slashes the error rate in spoken language identification by over 50%, highlighting the power of discriminative representation learning.