cluesurf-talk 2.0.0

Tokenized IPA for AI and NLP, a keyboard-friendly ASCII phonetic encoding that converts to and from IPA with one token per effective sound.

talk

A phonetic encoding. Convert between:

• IPA — the International Phonetic Alphabet (tʰa)
• talk — a plain-ASCII spelling of the same sounds (th~a)
• simple — a readable, human-facing rendering
• token — one Hangul code point per sound, for fixed-width machine use

Everything is derived from three data files and a double-array trie scan. Zero runtime dependencies.

This is the Python port of @cluesurf/talk. Both ports share one source of truth for the phonetic data.

Install

pip install cluesurf-talk
# or
uv add cluesurf-talk

Use

import talk

talk.ipa_to_talk("tʰa")    # "th~a"
talk.talk_to_ipa("th~a")   # "tʰa"
talk.readable("th~a")      # "tʰa"
talk.machine("th~a")       # one Hangul code point per sound

# Break a talk string into sounds with their features.
for sound in talk.segment("th~a"):
    print(sound.talk, sound.kind, sound.base and sound.base.talk,
          [m.feature for m in sound.modifiers])

# The full canonical inventory.
talk.enumerate_sounds()

API

Function	Description
ipa_to_talk(text)	IPA → talk
talk_to_ipa(text)	talk → IPA
readable(text)	talk → simple readable form
machine(text)	talk → one Hangul code point per sound
machine_outputs(text)	talk → list of per-sound code points
segment(text) / tokenize(text)	talk → list of Sound
enumerate_sounds()	the full canonical sound inventory
combine(base_talk, mods)	base + modifiers → canonical talk spelling

License

MIT

ClueSurf

Part of the ClueSurf toolset.