Original on Kaggle

%%capture
import os
os.chdir('/tmp')
!wget https://github.com/MontrealCorpusTools/Montreal-Forced-Aligner/releases/download/v1.0.1/montreal-forced-aligner_linux.tar.gz
!tar zxvf montreal-forced-aligner_linux.tar.gz
!ln -s /tmp/montreal-forced-aligner/lib/libpython3.6m.so.1.0 /tmp/montreal-forced-aligner/lib/libpython3.6m.so

os.chdir('/kaggle/working')
os.environ['LD_LIBRARY_PATH'] = f'{os.environ["LD_LIBRARY_PATH"]}:/tmp/montreal-forced-aligner/lib/'
os.environ['PATH'] = f'{os.environ["PATH"]}:/tmp/montreal-forced-aligner/bin/'

%%capture
!apt-get -y install libgfortran3

!mkdir /tmp/example

The example below is from section 488 (p. 239) of Gaeilge Chorca Dhuibhne by Diarmuid Ó Sé.

The provided transcription is: əs kiːn′ l′əm nə ˈheːn′ɪ v′eh ə bɪn′t′ vuːn er′

%%writefile /tmp/example/test1.lab
is cuimhin liom na haoinne a bheith ag baint mhóna air

Writing /tmp/example/test1.lab

Ibid, Section 488, p. 238

xuːərˈdiːs ˈgax ɑːt′

%%writefile /tmp/example/test2.lab
chuardaíos gach áit

Writing /tmp/example/test2.lab

MFA insists on having .wav files, which it reads, even though it makes no use of them for G2P

%%capture
!apt-get -y install sox

!sox -n -r 16000 -b 16 -c 1 -L /tmp/example/test1.wav trim 0.0 6.000
!sox -n -r 16000 -b 16 -c 1 -L /tmp/example/test2.wav trim 0.0 6.000

!mfa_generate_dictionary ../input/train-irish-mfa-model-fuaimeanna/g2p-munster.zip /tmp/example/ output

Setting up corpus information...

!cat output

cuimhin	k ɪ vˠ nʲ
áit	ɑː tʲ
a	ə
mhóna	vˠ oː n̪ˠ ə
bheith	vʲ ɛ
baint	bˠ ɪ nʲ tʲ
air	a ɾʲ
is	ɪ ʃ
na	n̪ˠ ə
gach	ɡ ə x
chuardaíos	x uə ɾˠ d̪ˠ iː ʌ sˠ
ag	a ɡ
liom	lʲ ʌ mˠ
haoinne	ɪ nʲ ɛ
Word Pronunciation Alt. Transcript Generated Correct? In context? Rule/Reason
is əs əsˠ (~ ɪʃ) ɪ ʃ ✔️ ❌ Exception: ios but correct before a slender consonant
cuimhin kiːn′ kiːnʲ k ɪ vˠ nʲ ❌ ❌ Missing grapheme: uimhi
liom l′əm lʲəmˠ (~ lʲʌmˠ) lʲ ʌ mˠ ✔️ ✔️ (See, e.g., section 291: l′um)
na nə n̪ˠə n̪ˠ ə ✔️ ✔️
haoinne ˈheːn′ɪ heːnʲɪ ɪ nʲ ɛ ❌ ❌
a ə ə ✔️ ✔️
bheith v′eh vʲɛ(h) vʲ ɛ ✔️ ❌ Section 9: h → ∅ / _ # -V
ag ə ə (~ ɪɟ) a ɡ ❌ ❌ ɪg′, section 60
baint bɪn′t′ bˠɪnʲtʲ bˠ ɪ nʲ tʲ ✔️ ✔️
mhóna vuːn vˠuːn̪ˠ(ə) vˠ oː n̪ˠ ə ✔️ ❌ ó → oː ~ uː / _ [+nasal], ə → ∅ / _ #
air er′ eɾʲ a ɾʲ ❌ ❌ Exception: eir
chuardaíos xuːərˈdiːs xuəɾˠd̪ˠiːsˠ x uə ɾˠ d̪ˠ iː ʌ sˠ ❌ ❌ Missing grapheme aío
gach ˈgax ɡax (~ ɡəx) ɡ ə x ✔️ ✔️ See section 810
áit ɑːt′ ɑːtʲ ɑː tʲ ✔️ ✔️