Skip to main content

Table 2 Experimental configurations

From: Phone lattice reconstruction for embedded language recognition in LVCSR

Backend
ID
Frontend
type
Lattice reconstruction algorithm
for VSM training and test
A0 LVCSR None (using word-based phone lattices)
A1 LVCSR Time-aligned phone lattice
A2 LVCSR Phoneme-expanded 1-best lattice
A3 LVCSR Frame-expanded N-best lattice
A3' LVCSR Pruned frame-expanded N-best lattice
PRVSM Phone-loop None (standard PRVSM system)
PRVSM-alt Phone-loop None (standard PRVSM system with LVCSR acoustic model)
  1. Each experiment was implemented twice, once with an English LVCSR frontend and once with a Mandarin LVCSR frontend