Somali ASR (Whisper)

This is a Whisper-based Automatic Speech Recognition (ASR) model fine-tuned on Somali language using the 8 Hours dataset. The model converts Somali speech to text.

🧠 Model Details

• Language: Somali (so)
• Trained on: 8 hours of Somali speech data
• Framework: 🤗 Transformers

📈 Key Metrics:

• Final WER: 0.231513
• Final Training Loss: 0.042500
• Final Validation Loss: 0.2177218

🗣️ Usage

To use the model, simply load it via the Hugging Face pipeline:

from transformers import pipeline

# Load the Somali ASR model
asr = pipeline("automatic-speech-recognition", model="hamaada/whisper-finetuned-somali-stt")

# Transcribe audio file to text
result = asr("path_to_somali_audio.wav")

# Output the transcribed text
print(result['text'])