Greek MFA Project

Description

This repository contains various scripts used to pre-process data, and run a forced alignment pipeline using the Montreal Forced Aligner for a Greek Alignment project.

File Description and Overview

• align.py

This script contains the entire pipeline. And will initiate all of the other scripts found inside this repository. This script requires two inputs: the data folder containing the audio and TextGrid files, and the tier that the Montreal Forced Aligner will use. python3 align.py /path/to/dataset/ nameOfTargetTier

• combine.py

This script will combine the original TextGrid file with the MFA aligned TextGrid file.

• extractTiers.py

This script will create a TextGrid file containing one tier for use with the Montreal Forced Aligner.

• posTag.py

This script uses the Spacy library to do part of speech tagging on the sentence tier and align the morphological tags with the word tier.

• punctFix.py

This file contains various data-cleaning commands. This file can be edited to further clean TextGrid files.

• wavPreProcess

This is a Praat script written by E. Chodroff that will re-sample .wav files to 16kHz and extract a single channel as required for the Montreal Forced Aligner.

• greek_acoustic_model.zip

This is the model that the Montreal Forced Aligner will use to align the transcripts to the audio files. This model was trained using the Greek Internet Corpus which contains ~72,000 tokens.

• phonDict.txt

This is the dictionary with orthographic to phoneme mappings.