alecristia
commented
6 years ago the next step is to create a phonological version to count syllables to do so, we could rely on the phonemizer package: https://github.com/bootphon/phonemizer For English, the FESTIVAL option will return syllable boundaries as ";esyll"
$ echo "hello world" | phonemize -l en-us-festival
An alternative for languages where only vowels are nuclei is to use the following code: https://github.com/laiafr/SegCatSpa/blob/master/Analysis_Pipeline/CatalanSpanish_recipe/catspa-syllabify-corpus.pl
it requires a list of vowels, which need to be provided for each language but my guess is that for both tseltal and spanish this list will do: aeiou
It also requires a list of "permissible onsets". The way to generate this is by:
- get all the words in the corpus (eg tr ' ' 'n' | sort | uniq)
- remove everything following the first vowel in the word (eg sed 's/[aeiou].*//g')
MarvinLvn
Implementing the following pipeline :
[.eaf] -> [csv. With onset, offset, ortho from.eaf, speakertier] -> [cleanup for .csv] -> [word counts + syllable counts]
Okko’s matlab scripts can be used to do current cleanup.