classla/wav2vec2-xls-r-parlaspeech-hr
🧠 AI Modelclassla
A high-performance Croatian automatic speech recognition model fine-tuned on the ParlaSpeech-HR dataset.
The classla/wav2vec2-xls-r-parlaspeech-hr model represents a significant advancement in low-resource language processing for Croatian. By utilizing the XLS-R architecture—a large-scale cross-lingual speech representation model—the authors have adapted the framework to handle the specific phonetic and linguistic nuances of Croatian parliamentary speech. The model is built on the PyTorch ecosystem and is fully compatible with the Hugging Face Transformers library, ensuring ease of deployment for production environments. It supports safetensors for secure and efficient model loading. This model is particularly effective for transcribing formal and semi-formal spoken Croatian, making it an essential resource for digital humanities, accessibility tools, and automated transcription services in the region.
💡Highlights
- ├─Fine-tuned XLS-R architecture
- ├─Optimized for Croatian language
- └─Transformers library compatible
🎯For
- ├─Computational linguists
- ├─Software developers
- └─Accessibility researchers