JONAHKYAGABA/Amharic-ASR-Model-Evaluation-Qualitative-

Amharic ASR Model Evaluation (Qualitative) # 🗣️ Amharic ASR Model Evaluation (Qualitative) This project evaluates **Automatic Speech Recognition (ASR)** models on Amharic audio datasets using Hugging Face's `pipeline`. It provides **qualitative feedback** through detailed CSVs showing predictions, references, and error metrics (WER/CER). --- ## 🎯 Project Objectives - Load one or more ASR models from Hugging Face - Evaluate on Amharic-language test sets - Compute **Word Error Rate (WER)** and **Character Error Rate (CER)** - Save results in model-wise CSV reports - Support batch inference and qualitative diagnostics ---