espnet2.text.cleaner.TextCleaner

source

class espnet2.text.cleaner.TextCleaner(cleaner_types: Collection[str] | None = None)

Bases: object

Text cleaner.

This class provides various text cleaning functionalities based on specified cleaner types. It supports different cleaning methods including tacotron, jaconv, Vietnamese, Korean, and whisper text normalization.

cleaner_types

A list of cleaner types to be applied.

• Type: list

• Parameters:cleaner_types (Optional *[*Collection *[*str ] ]) – A collection of cleaner types. If None, an empty list is used. It can also be a single string.

• Returns: The cleaned text after applying the specified cleaners.

• Return type: str

• Raises:

  • RuntimeError – If an unsupported cleaner type is specified or if the
  • Vietnamese cleaner is requested but not available. –

Examples

>>> cleaner = TextCleaner("tacotron")
>>> cleaner("(Hello-World);   &  jr. & dr.")
'HELLO WORLD, AND JUNIOR AND DOCTOR'

NOTE

Make sure to install required dependencies for all cleaner types to work properly, especially for Vietnamese cleaning.