search

xinjli / allosaurus

Allosaurus is a pretrained universal phone recognizer for more than 2000 languages
GNU General Public License v3.0
571 stars 88 forks source link

Is the output phone or phonemes? #55

Closed raotnameh closed 2 years ago

raotnameh commented 2 years ago

@xinjli Just curious is the output a list of phones or phonemes? Two cases:

  1. lang_id token provided
  2. lang_id token NOT provided
raotnameh commented 2 years ago

@xinjli There are 44 phonemes in English given the IPA list.

When I run the command to get a list of phones. they are 67. python -m allosaurus.bin.list_phone --lang eng

How can I map the phones to phonemes? In the paper, you mention an allophone layer. How do I get the phoneme as an output?

xinjli commented 2 years ago

Hi

Sorry for the late reply. The model currently only supports the output at the phone layer, the phoneme output is not supported yet. However, you can always map the phone output to phoneme easily by looking up a phone/phoneme (allophone) dictionary. For example, you can use the allovera mapping https://github.com/dmort27/allovera or phoible mapping https://phoible.org/inventories/view/160