Korean Speech Tools

This package contains a variety of tools for Korean speech and language processing, including:

• Hangul romanization.
• Jamo conversion.
• Grapheme-to-phoneme conversion (G2P).

It is mostly a collection of previously existing libraries that are unmaintained or include unnecessary dependencies that have been removed here. Refer to the Credits section below for details.

Installation

pip install ko-speech-tools

The core part of this package does not have any external dependencies. For G2P, the mecab-ko package is required, which can be installed with:

pip install ko-speech-tools[g2p]

Usage

Romanization:

>>> from ko_speech_tools import hangul_romanize

>>> hangul_romanize("물엿")
'mul-yeos'

Jamo conversion (documentation):

>>> from ko_speech_tools.jamo import h2j, j2hcj, j2h

>>> h2j('한굴')
'한굴'

>>> j2hcj(h2j('한굴'))
'ㅎㅏㄴㄱㅜㄹ'

>>> j2h('ㅇ', 'ㅕ', 'ㅇ')
'영'

G2P:

>>> from ko_speech_tools import G2p

>>> g2p = G2p()
>>> g2p("것입니다")
'거심니다'

Credits

This package combines and adapts the following packages:

• g2pkk: https://github.com/harmlessman/g2pkk (Apache-2.0), a fork of https://github.com/Kyubyong/g2pK
• hangul_romanize: https://github.com/youknowone/hangul-romanize (BSD-2-Clause)
• jamo: https://github.com/jdongian/python-jamo (Apache-2.0)

It additionally uses code from https://github.com/keithito/tacotron (MIT) to read CMUdict data.

The respective code can be used under the original license, see the individual files for details. Any new code is made available under Apache-2.0.