Metraipolis › archive › dc_6c8d45e50819
SpeakPay: Domain-Adaptive LoRA Fine-Tuning of Whisper for Low-Resource Nepali Financial Speech Recognition — arXiv:2609.01737
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.01737
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T04:16:29.540Z
- SHA-256 of the archived text
6dbd21477e33d2e0cb7b0a2fe8c194e670799b00b5a746a9287e8cf92d89b755- Stable URL of this page
https://metraipolis.com/documento/dc_6c8d45e50819
How to cite
SpeakPay: Domain-Adaptive LoRA Fine-Tuning of Whisper for Low-Resource Nepali Financial Speech Recognition — arXiv:2609.01737 — arXiv (en), https://arxiv.org/abs/2609.01737 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_6c8d45e50819 (Metraipolis archive dc_6c8d45e50819, sha256 6dbd21477e33d2e0cb7b0a2fe8c194e670799b00b5a746a9287e8cf92d89b755)
BibTeX
@misc{dc6c8d45e50819,
title = {SpeakPay: Domain-Adaptive LoRA Fine-Tuning of Whisper for Low-Resource Nepali Financial Speech Recognition — arXiv:2609.01737},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.01737},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_6c8d45e50819; sha256 6dbd21477e33d2e0cb7b0a2fe8c194e670799b00b5a746a9287e8cf92d89b755},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.01737. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Mobile payment applications in Nepal are graphically mediated and largely inaccessible to visually impaired users. This paper presents SpeakPay, a voice-first digital wallet, and documents the central technical contribution: a controlled study of domain adaptation for low-resource financial speech recognition. We introduce NepFinSpeech-403, a 403-utterance dataset of Nepali financial voice commands (send, load, and balance operations spanning 237 unique numerals), and fine-tune Whisper large-v2 with LoRA. On the held-out test set, the domain-adapted model reduces Word Error Rate from 129.95% (zero-shot baseline) to 42.58% --- a 67.2% relative reduction --- and improves Devanagari numeral recognition accuracy from 0.0% to 73.9%. We find that word-level metrics understate the practical task-level impact: domain adaptation improves the Transaction Success Rate from 1.67% to 33.33%, a roughly 20x gain. The improvement is consistent at the individual-utterance level (sign test, $p < 10^{-17}$) and across all command types. A data efficiency analysis shows that as few as 100 domain-specific utterances are sufficient to halve the zero-shot WER, with performance plateauing around 300 examples. Error analysis reveals systematic numeral confusion patterns (zero insertion/deletion, prefix hallucination) that account for the majority of remaining transaction failures. The trained system is deployed as a publicly accessible voice-first web application. All code, dataset, model weights, and this paper are released at this https URL .
Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-03 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.01737 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-17T03:26:49.755Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-03?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).