Search papers, labs, and topics across Lattice.
1
0
3
Counterintuitively, for low-resource Bengali speech processing, targeted data augmentation and heuristic post-processing outperform raw data scaling and extensive model retraining for ASR and speaker diarization, respectively.
