low-resource-asr

🎙️ PolyWhisper v9: efficient multilingual Indic ASR — frozen Whisper-Small + per-language LoRA for Hindi, Tamil, Telugu, Bengali, Marathi. Bengali −34.5%, Marathi −43.2% vs baseline. ONNX CPU-ready. Apache-2.0.

Kikuyu ASR - speech recognition for Kikuyu using MMS-300M, QLoRA, KenLM, and ONNX

Code-switching ASR adaptation for strong multilingual speech recognition models. Synthetic CSW data generation, Whisper adaptation, Bayesian LoRA (BLoRA), and robust multilingual ASR evaluation.

Contributions to AmericasNLP Shared Task for Low-Resource ASR