Dataset · MohamedRashad
ASR audio dataset with mixed Arabic-English speech
MohamedRashad/arabic-english-code-switching
Speech recordings with Arabic-English code-switching for training speech recognition systems.
What's inside
The dataset contains 10,000 to 100,000 audio and text pairs in Arabic and English for ASR tasks. It is built upon the arzen-llm-speech-ds dataset and augmented with cleaned and deduplicated fragments from various YouTube videos, doubling the initial data volume.
What you can do with it
- Fine-tuning ASR models (e.g., Whisper) for recognizing mixed Arabic-English speech
- Evaluating transcription quality under frequent language code-switching
- Tasks
- Automatic speech recognition
- Languages
- Arabic, English
- Size
- 10K–100K rows
- Downloads this month
- 395
- Likes
- 38
Models trained on it
- mohammedaly22/whisper-large-v3-turbo-egyptian-code-switchingAutomatic speech recognition 828
- MohamedRashad/Arabic-Whisper-CodeSwitching-EditionAutomatic speech recognition 370
- IbrahimAmin/code-switched-egyptian-arabic-whisper-smallAutomatic speech recognition 326
- mradermacher/MasriSwitch-Gemma3n-Transcriber-v1-i1-GGUF 309
- mradermacher/MasriSwitch-Gemma3n-Transcriber-v1-GGUF 297
- mohammedaly22/Metro-ASR-SmallAutomatic speech recognition 72
- oddadmix/MasriSwitch-Gemma3n-Transcriber-v1Automatic speech recognition 43
- MostafaKhidr/whisper-Large-ArabicEnglishAutomatic speech recognition 28
How to load it
from datasets import load_dataset
ds = load_dataset("MohamedRashad/arabic-english-code-switching")