Skip to content
AI News HubLIVE
Original source2 min read

SpeakPay: Domain-Adaptive LoRA Fine-Tuning of Whisper for Low-Resource Nepali Financial Speech Recognition

Summary

SpeakPay is a voice-first digital wallet that addresses graphical interface accessibility barriers in Nepal's mobile payment ecosystem. The paper introduces NepFinSpeech-403, a 403-utterance Nepali financial voice command dataset spanning send, load, and balance operations with 237 numerals, and applies LoRA-based domain adaptation to Whisper large-v2. This adaptation cuts Word Error Rate from 129.95% to 42.58% (67.2% relative reduction), raises Devanagari numeral accuracy from 0% to 73.9%, and improves Transaction Success Rate from 1.67% to 33.33% — roughly a 20x gain. As few as 100 domain utterances halve zero-shot WER, and performance plateaus around 300.

SourcearXiv Computational LinguisticsAuthor: Biraj Subedi
SpeakPay: Domain-Adaptive LoRA Fine-Tuning of Whisper for Low-Resource Nepali Financial Speech Recognition
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[Submitted on 1 Sep 2026]

Title:SpeakPay: Domain-Adaptive LoRA Fine-Tuning of Whisper for Low-Resource Nepali Financial Speech Recognition

View a PDF of the paper titled SpeakPay: Domain-Adaptive LoRA Fine-Tuning of Whisper for Low-Resource Nepali Financial Speech Recognition, by Biraj Subedi

View PDF HTML (experimental)

Abstract: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

new | recent | 2026-09

Change to browse by:

cs

References & Citations

NASA ADS

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • Introduces NepFinSpeech-403, an open dataset of 403 Nepali financial voice commands covering send, load, and balance operations with 237 unique numerals.
  • Domain-adaptive LoRA fine-tuning of Whisper large-v2 reduces WER from 129.95% zero-shot to 42.58%, a 67.2% relative improvement.
  • Transaction Success Rate improves roughly 20x, from 1.67% to 33.33%, with consistent gains across all command types.
  • Data efficiency analysis shows 100 domain utterances are enough to halve zero-shot WER, with gains plateauing near 300 examples.

Highlights and analysis are generated automatically and may contain errors. Check the original source.