Here is the proposed file hierarchy:

• <dataset>/
  • segments/
    • *.seg
  • acfeats/
    • *.slx2
  • <def>/
    • mcfeats/
      • *.ling
    • models/ (*.model files)
      • 1v1/
      • <language_set>/
    • results/
      • <language_set>/
      • <language_pair>/
    • ling.batch
    • def.xml
  • wav.batch
  • train.dataset
  • devtest.dataset
  • eval.dataset

(The bracketed items are variables)

List of Batch Files:

Type	Input	Output
Phone Recognizer	.wav	.seg
RecOptimizer	.wav,trans	-
PRAAT	.seg	.slx2
Feature Extraction + Conversion	.slx2	.ling
Learner	.ling	.model
Classifier	.ling	prob

First two are “per language” batch files, last two are “full batch”. Spoken Language ID

Back to top