G2PModel#

class montreal_forced_aligner.models.G2PModel(source, root_directory=None)[source]#

Bases: Archive

Class for G2P models

Parameters:

source (str, Path) – Path to source archive
root_directory (str, Path) – Path to save exported model

add_fst_model(source_directory)[source]#

Add FST file into archive

Parameters:: source_directory (str) – Source directory path

add_meta_file(g2p_trainer)[source]#

Construct metadata information for the G2P model from the dictionary it was trained from

Parameters:: g2p_trainer (G2PTrainer) – Trainer for the G2P model

add_sym_path(source_directory)[source]#

Add symbols file into archive

Parameters:: source_directory (str) – Source directory path

export_fst_model(destination)[source]#

Extract FST model path to destination

Parameters:: destination (str) – Destination directory

property fst_path#: G2P model’s FST path

property grapheme_sym_path#: G2P model’s grapheme symbols path

property meta#: Metadata for the G2P model

property sym_path#: G2P model’s symbols path

validate(word_list)[source]#

Validate the G2P model against a word list to ensure that all graphemes are known

Parameters:: word_list (Collection[str]) – Word list to validate against
Returns:: False if missing graphemes were found
Return type:: bool

validate_phone_symbols(dictionary)[source]#

Validate this G2P model against a pronunciation dictionary to ensure their phone sets are compatible (i.e., G2P model will not generate phones not in the dictionary)

Parameters:: dictionary (DictionaryMixin) – DictionaryMixin to compare phone sets with
Raises:: PronunciationAcousticMismatchError – If there are phones missing from the acoustic model