arXiv AI

Nepali-English Code-Mixed Speech Recognition: Data Preparation and Adaptation of a Speech Foundation Model

The paper presents a study on adapting the SraVaani speech recognition model to Nepali-English code‑mixed speech. Using a 172.8‑hour corpus across six Nepali domains, the authors fine‑tune the 449M‑parameter hybrid TDT‑CTC model, addressing tokenizer limitations for digits and symbols by generating Nepali and English readings and selecting plausible forms via zero‑shot ASR. Full fine‑tuning reduces the test word error rate from 48.6% to 17.1%, matching Whisper large‑v3‑turbo, while experiments on freezing the encoder and excluding numeric clips illustrate the impact of model components and data content.

arXiv Computation and Language
Oct 23

Synthetic-to-Real Code-Switched ASR: A Spontaneous Nepali-English Podcast Benchmark

The paper introduces a new benchmark of 13,044 spontaneous Nepali‑English code‑switched podcast utterances, totaling 15.45 h, and evaluates a two‑stage synthetic‑to‑real adaptation framework for ASR. Stage 1 uses supervised fine‑tuning on 0.95 h of synthetic speech, reducing Mixed Error Rate from 23.40 % to 21.52 % and switch‑point error rate from 35.89 % to 28.77 %. Stage 2 applies Group Relative Policy Optimization with script and transcription rewards, yielding modest but statistically significant further improvements, particularly in English script confusion and switch‑point accuracy. The study also shows that an auxiliary script reward term offers no benefit beyond MER‑only optimization in this setting, and that the synthetic SFT provides the main cross‑domain transfer to spontaneous human speech.

By Firoj Paudel
arXiv Computation and Language
Aug 24

Building and Evaluating a Synthetic Bengali Speech Resource for Telecom Customer Care

The paper introduces a synthetic Bengali speech dataset tailored for telecom customer‑care applications, comprising 10,000 audio‑text pairs (≈26.82 hours) with predefined train, validation, and test splits. The data were generated using OmniVoice voice‑cloning, and include both original and normalized transcripts for ASR/STT use. Automatic intelligibility evaluation with a fine‑tuned Whisper model shows an average WER of 2.54% and CER of 0.59%, indicating strong text‑audio consistency, while the authors note limitations of synthetic speech and STT‑based evaluation.

By Kawshik Kumar Paul, Md. Nafiul Alam Fuji
arXiv Computation and Language
Aug 27

Fine-Tuning Whisper for Automatic Speech Recognition in Baniwa: A Preliminary Study

The study fine‑tunes the Whisper Small model for Automatic Speech Recognition (ASR) in Baniwa, an indigenous Arawakan language. Using a 0.54‑hour corpus of 1,373 manually transcribed recordings, the fine‑tuned model achieved a Word Error Rate of 37.5% and a Character Error Rate of 7.45%. These results provide an initial baseline for Baniwa ASR and suggest that multilingual foundation models can be adapted to extremely low‑resource languages.

By Leonardo Duart, Tiago Fonseca, Thiago Chac\'on
arXiv Computation and Language
Sep 3

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

SpeakPay is a voice‑first digital wallet designed to make mobile payment apps in Nepal accessible to visually impaired users. The paper introduces NepFinSpeech‑403, a 403‑utterance Nepali financial voice command dataset, and demonstrates that fine‑tuning Whisper large‑v2 with LoRA reduces the Word Error Rate from 129.95% to 42.58% and improves Devanagari numeral recognition from 0.0% to 73.9%. Domain adaptation also boosts the Transaction Success Rate from 1.67% to 33.33%, with as few as 100 domain‑specific utterances halving the zero‑shot WER.

By Biraj Subedi
arXiv AI
Sep 4

Building and Evaluating Fixed-Voice Thai TTS from Synthetic Speech

The paper presents a method for creating a compact fixed‑voice Thai text‑to‑speech system by training a student model on synthetic speech generated from a large voice‑cloning teacher. By using only a short 15‑second voice reference and carefully filtering synthetic data, the authors build an 82‑million‑parameter model, Wayu‑Paxa‑TTS‑Edge, that runs on device without reference audio. The system achieves strong performance—68.2 % challenge‑set keyword accuracy, 91.4 % pause precision, and low character error rates—while outperforming its teacher and approaching the quality of a larger Gemini 3.1 model.

By Kunat Pipatanakul, Potsawee Manakul, Warit Sirichotedumrong, Sittipong Sripaisarnmongkol, Pakorn Nathong, Phatrasek Jirabovonvisut