Search papers, labs, and topics across Lattice.
LY Corporation
3
0
4
MOS models miss critical prosodic nuances, leading to a disconnect between human perception and automated assessments in speech quality.
Conventional speech quality models miss critical pitch-accent errors, but PASQA accurately ranks accent severity and aligns with human judgments.
Streaming grapheme-to-phoneme conversion for unsegmented languages is now possible thanks to a Conformer-CTC architecture that learns grapheme-phoneme alignments without needing explicit word boundaries.
