🎧 Assignment — Forced Alignment using Montreal Forced Aligner (MFA)

📌 Objective

To automatically align speech audio and text transcripts at the word and phoneme level using the Montreal Forced Aligner (MFA).

---

🧩 Steps to Run

# 1️⃣ Create and activate environment
conda create -n mfa_env -c conda-forge montreal-forced-aligner -y
conda activate mfa_env

# 2️⃣ Download models
mfa model download dictionary english_us_arpa
mfa model download acoustic english_mfa

# 3️⃣ Prepare dataset
# Ensure data/ready_corpus contains .wav and .txt pairs

# 4️⃣ Validate
mfa validate data/ready_corpus english_us_arpa english_mfa

# 5️⃣ Align
mfa align data/ready_corpus english_us_arpa english_mfa outputs/aligned

📊 Outputs

Alignment Files → outputs/aligned/*.TextGrid

Alignment Report → outputs/aligned/alignment_analysis.csv

Each .TextGrid contains:

Word tier → timestamps for words

Phone tier → timestamps for phonemes

🔍 Visualization

Open in Praat
:

Open → Read from file → F2BJ_RLP1.wav

Open → Read from file → F2BJ_RLP1.TextGrid

Select both → View & Edit

🧠 Observations

Word and phone boundaries aligned accurately.

Minor timing deviations in fast speech segments.

english_us_arpa dictionary and english_mfa acoustic model performed well.