KabulaBenjamin/kabras-asr-research

# Kabras ASR Research Research and analysis companion to the kabras-ai-project app. Computes Word Error Rate (WER), confidence statistics, and visualisations from the translation logs produced by the main app. Designed to support ongoing improvement of the Kabras language lexicon and ASR pipeline. ## Project structure ``` kabras-asr-research/ ├── analysis/ │ ├── stats.py # Summary statistics (match rate, avg confidence, WER) │ ├── visualize.py # Saves confidence & WER plots to analysis/plots/ │ └── plots/ # Auto-generated plot images ├── notebooks/ │ ├── wer_analysis.ipynb # Computes WER, writes results back to CSV │ └── confidence_plots.ipynb # Visualisations (run after wer_analysis) ├── datasets/ # Local research datasets (not tracked by git) ├── config.py # Central path configuration — edit LOG_PATH here ├── requirements.txt └── README.md ``` ## Setup ```bash git clone github.com cd kabras-asr-research python -m venv venv venv\Scripts\activate # Windows pip install -r requirements.txt ``` ## Configuration Open `config.py` and set `LOG_PATH` to point to the CSV written by the kabras-ai-project app: ```python LOG_PATH = r"C:\Users\YOU\MyProjects\kabras-ai-project\app\datasets\translations_log.csv" ``` ## Usage ### 1. Run the Kabras AI app first Make at least one translation at 127.0.0.1 to generate `translations_log.csv`. ### 2. Print statistics ```bash python analysis/stats.py ``` ### 3. Generate and save plots ```bash python analysis/visualize.py ``` Plots are saved to `analysis/plots/`. ### 4. Run notebooks (in order) ```bash jupyter lab ``` - Open `notebooks/wer_analysis.ipynb` → Run All - Open `notebooks/confidence_plots.ipynb` → Run All ## Data flow ``` kabras-ai-project app │ │ writes translations_log.csv on every translation ▼ kabras-asr-research wer_analysis.ipynb → adds WER + CalcConfidence columns to C …