Syllabifier for Luxembourgish

Based on the syllabifier from the Penn Phonetics Toolkit, this is an adaption for Luxembourgish.

The Python script syllabifer.py takes IPA-transcribed words of Luxembourgish as input and returns the segments grouped as syllables. The phonetic segments in the input have to be separated by spaces, e.g. æ p ə l t ɛːɐ t ɐ ɕ ɐ (Äppeltäertercher).

The script applies the Maximum Onset Principle to assign as many consonants to a potential syllable onset as long as this onset is permitted in the language. These allowed onsets are specified in luxembourgish.cfg (and also in the script itself) as onsets; permitted consonants and nuclei are specified in the configuration as well.

The lists with IPA transcribed words comes from the training data for my g2p system.

Although developed and tested for Luxembourgish, this system should also work in principle for IPA data for lots of other languages. It is only necessary to adapt the configuration for onsets, nuclei and consonsants.

Usage

The Python notebook Syllabifier provides a simple example.

import syllabifier
language = syllabifier.Luxembourgish

syllables = syllabifier.syllabify(language, "æ p ə l t ɛːɐ t ɐ ɕ ɐ")

# some test words
# æ p ə l t ɛːɐ t ɐ ɕ ɐ
# ɑ p ʀ o v i z i̯ o n ɜɪ ə ʀ ə n
# ʃ ɑ̃ː m b ɐ s p ʀ e z i d æ n t i n ə n
# k ɑ ʀ ɑ k t ə ʀ eː ʑ ə n ʃ ɑ f t ə n
# k ɑ ʀ i s m aː
# ʃ i m iː s p ʀ o f æ s oː ʀ i n ə n
# ɕ i ʀ u ʀ ʑ i n ə n
# k w ɑ f œː ʀ s s ɑ l ɑ̃ː
# d e k l ɑ ʀ ɑ ts i̯ əʊ n ə n
# d ʀ oː ɡ ə ʃ t ɑ t i s t i k

for stress, onset, nucleus, coda in syllables :
   print(" ".join(onset), " ".join(nucleus), " ".join(coda))

æ 
p ə l
t ɛːɐ 
t ɐ 
ɕ ɐ 

When run from a terminal:

python syllabifier.py luxembourgish.cfg < lux_ipa.txt > lux_ipa_syllabified.txt

Limitations

• Compounds like b l i n t d aː ʀ m o p e ʀ ɑ ts i̯ əʊ n where the m is falsely attributed to the next syllable, because m o is of course a permitted onset. To solve this, a morphological pre-segmentation would be necessary.

• Intervocalic single consonants after short vowel are ambisyllabic and should thus be attributed to both syllables, which is not possible in this framework. In ʃ t ʀ e k ə l the k is in the second, while in z ɑ ŋ ə n the ŋ goes to the first syllable.