LmCorpusTrainerMixin

class montreal_forced_aligner.language_modeling.trainer.LmCorpusTrainerMixin(**kwargs)

Bases: LmTrainerMixin, TextCorpusMixin

Top-level worker to train a language model from a text corpus

Parameters:

• order (int) – Ngram order, defaults to 3

• method (str) – Smoothing method for the ngram model, defaults to “kneser_ney”

• count_threshold (int) – Minimum count needed to not be treated as an OOV item, defaults to 1

See also

LmTrainerMixin

For language model training parsing parameters

TextCorpusMixin

For corpus parsing parameters

TopLevelMfaWorker

For top-level parameters

property cnts_path

Internal path to counts file

evaluate()

Run an evaluation over the training data to generate perplexity score

property far_path

Internal path to FAR file

property meta

Metadata information for the language model

property sym_path

Internal path to symbols file

train()

Train a language model

train_large_lm()

Train a large language model

property training_path

Internal path to training data