frhyme 0.3

Guess the last phonemes of a French word

frhyme -- a toolkit to guess the last phonemes of a French word Repository URL: https://gitlab.com/a3nm/frhyme Python package name: frhyme

== 0. Author and license ==

frhyme is copyright (C) 2011-2019 by Antoine Amarilli

frhyme is free software, distributed under an MIT license: see the file LICENSE for details of the licensing terms that apply to frhyme.

Many thanks to Julien Romero who maintains the PyPI package for frhyme.

The file "frhyme.json" in the directory "frhyme" is a derivative work of the French lexical database Lexique http://www.lexique.org/, version 3.83, by Boris New http://psycho-usmb.fr/boris.new/ and Christophe Pallier http://www.pallier.org/. Hence, this file is under the same license as Lexique, namely, the license CC BY SA 4.0 (according to the file README-Lexique.txt in the downloadable archive of Lexique). The license in LICENSE does not apply to this file "frhyme/frhyme.json".

== 1. Features ==

frhyme is a tool to guess what the last phonemes of a French word are. It is trained on a list of words with associated pronunciation, and will infer a few likely possibilities for unseen words using known words with the longest common prefix, using a trie for internal representation.

== 2. Installation ==

You need a working Python3 environment to run frhyme.

You can install frhyme directly with pip by doing:

pip3 install frhyme

You can also manually clone the project repository and use frhyme directly from there.

== 3. Usage ==

You can either run

frhyme.py [NBEST]

giving one word per line in stdin and getting the NBEST top pronunciations on stdout (default is 5), or you can import frhyme in a Python program and call frhyme.lookup(word, NBEST) which returns the NBEST top pronunciations (default is 5).

The pronunciations returned are annotated with a confidence score (the number of occurrences in the training data). They should be sensible up to the longest prefix of the input word that occurs in the training data, but they may be prefixed by garbage.

The pronunciations are given in a variant of X-SAMPA which ensures that each phoneme is mapped to exactly one ASCII character: the substitutions are "A~" => "#", "O~" => "$", "E~" => ")", "9~" => "(".

== 4. Training ==

This section explains how the file "frhyme.json" can be prepared. You do not need to do this to use frhyme, but it can be useful if you want to create a pronunciation database from a different source.

The provided "fryhme.json" file was trained on a custom variant of the database Lexique http://www.lexique.org/, with some additions. You can regenerate it as follows:

git clone 'https://a3nm.net/git/lexique' cd scripts ./make.sh 4 <(cut -f 1,2 ../lexique/lexique_my_format | uniq) additions > ../frhyme/frhyme.json

The value "4" indicates the number of trailing phonemes to keep, and can be changed. Beware, this process can take up several hundred megabytes of RAM. The resulting file should be accurate on the French words of Lexique.