🧪 Test?View on arXiv
SpeakPay: Domain-Adaptive LoRA Fine-Tuning of Whisper for Low-Resource Nepali Financial Speech Recognition
Subedi Biraj, Author 2, Author 3, Author 4, Author 5
fine-tuningspeech recognitionlow-resourcefinancial technology
2609.01737
Builder Relevance
2h ago80%
Abstract
This paper presents SpeakPay, a voice-first digital wallet aimed at improving accessibility for visually impaired users in Nepal through domain-adaptive speech recognition.
Reality Card
Core Claim
The domain-adapted model reduces Word Error Rate from 129.95% to 42.58% and improves Transaction Success Rate from 1.67% to 33.33%.
Method / Result
67.2% relative reduction in Word Error Rate.
Limitations
Systematic numeral confusion patterns remain a challenge, affecting transaction success.
Paper to code
Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.
No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.