lao-text-evaluation 0.2.1

Lao OCR Evaluation Tool with CER/WER and visual alignment

Lao Text Evaluation 🧪🇱🇦

Grapheme-aware evaluation toolkit for Lao OCR or text models — including CER, WER, character-level error breakdowns, and alignment visualizations.

---

✨ Features

• ✅ Character Error Rate (CER) and Word Error Rate (WER)
• ✅ Grapheme-level alignment using Unicode-aware splitting
• ✅ Lao-specific character classification: consonants, vowels, tone marks
• ✅ Visual alignment plots (color-coded matches, insertions, deletions, substitutions)
• ✅ CLI interface for single or batch file comparison
• ✅ Export results to CSV
• ✅ Generate sample data for quick testing

---

🛠 Installation

pip install lao-text-evaluation

---

🚀 CLI Usage

🔹 Generate Sample Data

Quickly create example ground truth and prediction files for testing:

python -m lao_text_evaluation.cli --generate-sample-data

• This will create sample/data/gt and sample/data/pred folders with example .txt files.

🔹 Single Pair Evaluation

python -m lao_text_evaluation.cli \
    --gt "ນ້ຳໃຈ" \
    --pred "ນ້ຳໃສ່" \
    --filename "sample1" \
    --plot \
    --plot-dir ./plots

🔹 Batch Folder Mode

python -m lao_text_evaluation.cli \
    --gt-path sample/data/gt \
    --pred-path sample/data/pred \
    --plot-sample 5 \
    --plot-dir ./plots \
    --save-csv results.csv \
    --append-csv

• .txt files must match by filename between ground truth and predictions.
• Plot images (optional) will be saved to --plot-dir.

---

🧪 Output Example

• ✅ CSV format: Filename, CER, WER, Removed_*, Inserted_*, Replaced Graphemes
• 🖼️ Visual plot: character alignment with color-coded errors

---

🧩 Module Usage (Python API)

from lao_text_evaluation.metrics import compute_cer, analyze_grapheme_errors
compute_cer("ນ້ຳໃຈ", "ນ້ຳໃສ່")

---

📂 Folder Structure

lao_text_evaluation/
├── metrics.py
├── utils.py
├── plotting.py
├── config.py
├── cli.py

---

📜 License

MIT License © 2025 Khonepaseuth Sounakhen