emPhon

Phonetic transcriber for the xtsv framework.

Requirements

• Python >= 3.6
• make

Installation

• Clone the repository: git clone https://github.com/levaid/emPhon
• make build
• pip install dist/*.whl

Usage

• Same as any other module using the xtsv framework - either as part of the emtsv framework or as separate module.
• The module needs form and anas fields and produces the phon field and by default, it prepends the phonetic form of the entire sentence

Configurations

The module takes command line arguments. By default, the module produces IPA output in strict xtsv format.

Command line arguments:

• --ipaize or --no-ipaize toggles IPA-ization, it produces the inner representation which uses exactly one unicode character for each phoneme. Default: on.
• --opt-palatal-assim or --no-opt-palatal-assim toggles optional palatal assimilation for the t/d+ny clusters, e.g. lapátnyél -> lapátynyél. Default: off.
• --include-sentence or --no-include-sentence toggles the inclusion of the entire phonetic form as a comment before each sentence. Default: on.

Example output

# phon = ɒ mɛɡoldaːʒbɒ ɒkaːr moʃt meːɡ nɛm iʃ ɛŋɡɛdeːjɛzɛtː oltaːʃok iʃ bɛjaːt͡shɒtnɒk .
A       ɒ
megoldásba      mɛɡoldaːʒbɒ
akár    ɒkaːr
most    moʃt
még     meːɡ
nem     nɛm
is      iʃ
engedélyezett   ɛŋɡɛdeːjɛzɛtː
oltások oltaːʃok
is      iʃ
bejátszhatnak   bɛjaːt͡shɒtnɒk
.       .

Paper

Can be downloaded from here. Please cite the following paper using this module:

@InProceedings{   Kulcsar:2021,
  author        = {Virág Kulcsár and Dániel Lévai},
  title         = {em{P}hon: Morphologically sensitive open-source phonetic transcriber},
  booktitle     = {{XVII}. Conference on Hungarian Computational Linguistics  ({MSZNY}2021)},
  year          = 2021,
  address       = {Szeged}
}

Miscellaneous

If different IPA (or transcription) is needed, the emphon/ipa_key.json file contains the key, which is human-modifiable.

License

This module is licensed under the LGPL 3.0 license.