PyniniValidator

class montreal_forced_aligner.g2p.trainer.PyniniValidator(word_list=None, **kwargs)

Bases: PyniniGenerator, TopLevelMfaWorker

Class for running validation for G2P model training

Parameters:

word_list (list[str]) – List of words to generate pronunciations

See also

PyniniGenerator

For parameters to generate pronunciations

compute_validation_errors(gold_values, hypothesis_values)

Computes validation errors

Parameters:

• gold_values (dict[str, set[str]]) – Gold pronunciations

• hypothesis_values (dict[str, list[str]]) – Hypothesis pronunciations

property data_directory

Data directory

property data_source_identifier

Dummy “validation” data source

evaluate_g2p_model(gold_pronunciations)

Evaluate a G2P model on the word list

Parameters:

gold_pronunciations (dict[str, set[str]]) – Gold pronunciations

property evaluation_csv_path

Path to working directory’s CSV file

setup()

Set up the G2P validator

property words_to_g2p

Words to produce pronunciations