BaltiVoice: A Speech Corpus and Fine-tuned Whisper ASR System for the Balti Language
Domaine:
natural language processingType de record:
paperdatasetmodel
Créateur:
Hôte:
We present BaltiVoice, a 16.8-hour read-speech corpus for Balti (ISO 639-3: bft), a Tibetic language spoken in Gilgit-Baltistan, Pakistan, with no prior publicly available ASR resources. The corpus contains 10,060 validated utterances in native Nastaliq script, derived from Mozilla Common Voice recordings. Fine-tuning OpenAI Whisper-small yields a Word Error Rate (WER) of 26.74% and a Character Error Rate (CER) of 8.67% on a 538-utterance speaker-disjoint validation set, down from a zero-shot baseline of 159.19% WER and 152.52% CER. A Whisper-base fine-tuned on the same data achieves 44.54% WER and 15.61% CER, confirming that model capacity matters for this low-resource setting. The dataset, fine-tuned model, and a live transcription demo are publicly available on HuggingFace. 6 pages, 3 figures, 4 tables. Code and data available at mohdali-dev/BaltiVoice-ASR
Similaires
AV55CS/Speech-to-Text-Fine-tuning-Whisper-Tiny-for-Swahili-ASRFine-Tuned Self-Supervised Speech Representations for Language Diarization in Multilingual Code-Switched SpeechAnass-Srk/fine-tuned-whisper-medium-darijaFine-tuning Whisper Tiny for Swahili ASR: Challenges and Recommendations for Low-Resource Speech RecognitionLOUA158/mms-fulfulde-asr-fine-tuned-fulJONAHKYAGABA/-Whisper-ASR-Fine-Tuning-for-Amharic