Commission_Viens_Dataset

Installation

python -m virtualenv .env
source .env/bin/activate
pip install -r requirements.txt

python scraper.py

Use the Montreal Forced Aligner to align audio with text (only one file):

mfa align_one audiofile.wav audiotext_to_align.txt french_mfa french_mfa test --use_mp --beam 100 --retry_beam 400

to align a folder:

mfa align corpus/path/ french_mfa french_mfa output/path --use_mp --beam 10000 --retry_beam 40000

(beam and retry_beam must be adjusted according to the duration of the audio file.)

for aligning only one file

python read_mfa.py --text_file input/path/filename.txt --textgrid_file input/path/filename.TextGrid --output_path output/corpus

for aligning file pairs from a folder (filename.txt, filename.TexGrid)

python align_from_textgrid --folder input/path/ --output_path output/corpus

python generate_segments_from_json.py file