JONAHKYAGABA/Afaan-data-prep-oromo

# Afaan-data-prep-oromo Afaan Oromo ASR & TTS Data Preparation This project prepares, cleans, and structures audio-text datasets in Afaan Oromo for use in automatic speech recognition (ASR) and text-to-speech (TTS) tasks. It includes data validation, audio-transcription alignment, normalization, vocabulary creation, and upload to the Hugging Face Hub. 📌 Key Features 🧹 Clean and validate metadata.csv for text-audio alignment 🔗 Match .wav files with their corresponding transcripts 🎧 Format audio using Hugging Face's datasets library 🔤 Normalize transcriptions for consistency 📈 Filter outliers and limit audio lengths for model training ☁️ Upload final DatasetDict to Hugging Face Hub 🧠 Generate vocabulary files (vocab.json) for tokenizer training