Skip to content

Home

Jump to bottom
Hywel Stoakes edited this page Dec 5, 2017 · 4 revisions

A Pan-Australian MAUS model.

Welcome to the supplementary material for a Pan-Australian forced alignment model. This is a very early prototype which may have mixed results for some Australian languages. More data is required, both for testing and for re-training the model. In the meantime we are very interested to hear about your progress.

Three Steps to aligned data:

  • Step 1: The grapheme to phoneme rules
  • Step 2: Chunking the data so as to not overwhelm MAUS
  • Step 3: Running the Pan-Australian MAUS model.

  • Step 4: Correct the alignment.
  • Step 5: Contribute your labelled files back so we can train a further forced alignment (or ideally automatic recognition) model.

© Hywel Stoakes 2017, The University of Melbourne.

Home

Clone this wiki locally