non-speech annotations in version 2.0.0rc3

[Jeff Mielke]

Hi. I'm using mfa version 2.0.0rc3 to align with a transcript that includes {LG} and {SL}. Previously I've used dictionary entries {LG} spn and {SL} sil, but when I include these in my dictionary now and align using the pretrained english acoustic models, I receive this message during validation:

  Acoustic model compatibility
  ============================

    2 phones not in acoustic model
    10 ignored pronunciations

    Phones missing acoustic models:

        sil and spn

If I use the default english dictionary or another dictionary that doesn't have the {LG} and {SL} entries, {LG} and {SL} get treated as oovs because they're not in the dictionary. If I include them in the dictionary, they get treated as oovs because they refer to missing acoustic models. So I can't figure out how to use non-speech annotations.

thanks

Jeff Mielke