55 lines
2.6 KiB
Markdown
55 lines
2.6 KiB
Markdown
---
|
|
language:
|
|
- ar
|
|
license: apache-2.0
|
|
base_model: openai/whisper-small
|
|
datasets:
|
|
- mozilla-foundation/common_voice_11_0
|
|
model-index:
|
|
- name: Whisper-small-ar
|
|
results: []
|
|
library_name: transformers
|
|
pipeline_tag: automatic-speech-recognition
|
|
---
|
|
|
|
# Arabic-Whisper Small
|
|
|
|
## Description
|
|
|
|
Whisper-small-ar is an Automatic Speech Recognition (ASR) model fine-tuned specifically for the Arabic language using the Whisper model architecture. ASR models are designed to convert spoken language into written text. This model has been fine-tuned on the Mozilla Common Voice dataset (version 11.0) to transcribe spoken Arabic speech into textual form.
|
|
|
|
### Key Features
|
|
|
|
- **Arabic Language Support:** Whisper-small-ar is optimized for recognizing and transcribing the Arabic language accurately. It can handle various Arabic dialects and accents.
|
|
|
|
- **Transformer Architecture:** The model is built on a powerful Transformer-based encoder-decoder architecture, which has demonstrated state-of-the-art performance in various natural language processing tasks, including ASR.
|
|
|
|
- **Fine-tuned for Arabic ASR:** The model has undergone a fine-tuning process on a substantial amount of Arabic speech data, making it well-suited for a wide range of ASR applications in Arabic, such as transcription of podcasts, call center recordings, and more.
|
|
|
|
- **Open-Source:** Whisper-small-ar is open-source and available for use by the research and developer community, facilitating the advancement of ASR technology for the Arabic language.
|
|
|
|
- **Compatible with Hugging Face Transformers:** You can easily integrate and utilize this model in your ASR projects using the Hugging Face Transformers library.
|
|
|
|
### Use Cases
|
|
|
|
Whisper-small-ar can be employed in a variety of ASR use cases, including:
|
|
|
|
- **Transcription Services:** Convert spoken Arabic content, such as audio recordings, podcasts, or videos, into written text for indexing, search, or translation purposes.
|
|
|
|
- **Voice Assistants:** Enhance voice-activated systems and virtual assistants with accurate Arabic speech recognition capabilities.
|
|
|
|
- **Language Processing Applications:** Integrate the model into applications involving Arabic language processing, such as sentiment analysis, keyword extraction, and more.
|
|
|
|
- **Multilingual ASR:** Combine Whisper-small-ar with other multilingual ASR models for applications requiring recognition of multiple languages.
|
|
|
|
## Usage
|
|
```python
|
|
# Use a pipeline as a high-level helper
|
|
from transformers import pipeline
|
|
|
|
pipe = pipeline("automatic-speech-recognition", model="ayoubkirouane/whisper-small-ar")
|
|
|
|
def transcribe(audio):
|
|
text = pipe(audio)["text"]
|
|
return text
|
|
``` |