mlsanityData to 19 Oct 2026

Responsible ASR: Overcoming Challenges of Foundational Models in Narrow-Band and Low-Resource Settings

arXiv 2606.18659 ↗First posted 17 Jun 2026cs.SDcited by 0

On the map: Speech and audio language models → Multilingual and low-resource ASR

Abstract (arXiv)

Telephony conversations worldwide are conducted over narrow-band channels and are often spontaneous and colloquial in nature. This paper evaluates the performance of widely used foundational automatic speech recognition (ASR) models -- both open-source and commercial -- on narrow-band conversations in Hindi, a low-resource language, and Indian-accented English, a low-resource accent. We first assess these models in a zero-shot setting and find that their performance remains suboptimal across the board. Highlighting the challenges faced by ASR models in narrow-band and low-resource language scenarios, we further investigate the impact of fine-tuning open-source models using a limited set of real-life annotated recordings. Our findings indicate that while fine-tuning provides some improvements, its effectiveness varies across languages and accents, largely influenced by the amount of data encountered during pretraining

Key facts. No citing papers in the map yet (data through 19 Oct 2026). Placed in Multilingual and low-resource ASR, one of 11 directions in Speech and audio language models. Builds on 14 of its 22 references that are in the map. First posted on arXiv 17 Jun 2026.

Builds on

14 of 22 references.

Closest ancestors · 3

Same direction, newest first

  1. 2023Vistaar: Diverse Benchmarks and Training Sets for Indian Language ASR24 May
  2. 2023Scaling Speech Technology to 1,000+ Languages22 May
  3. 2021XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale17 Nov

Other references · 11

  1. 2017Attention Is All You Needcited by 61,180
  2. 2022Robust Speech Recognition via Large-Scale Weak Supervisioncited by 4,661
  3. 2020wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representationscited by 4,385
  4. 2020Conformer: Convolution-augmented Transformer for Speech Recognitioncited by 2,293
  5. 2018SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processingcited by 2,060
  6. 2019SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognitioncited by 1,636
  7. 2020Pushing the Limits of Semi-Supervised Learning for Automatic Speech Recognitioncited by 211
  8. 2022Self-supervised Learning with Random-projection Quantizer for Speech Recognitioncited by 195
  9. 2020Improved Noisy Student Training for Automatic Speech Recognitioncited by 154
  10. 2024Open-Source Conversational AI with SpeechBrain 1.0cited by 68
  11. 2024Open Implementation and Study of BEST-RQ for Speech Processingcited by 18

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2021Towards Building ASR Systems for the Next Billion Users
  2. 2024ASR Benchmarking: Need for a More Representative Conversational Dataset
  3. 2025Are ASR foundation models generalized enough to capture features of regional dialects for low-resource languages?
  4. 2025Breaking the Transcription Bottleneck: Fine-tuning ASR Models for Extremely Low-Resource Fieldwork Languages
  5. 2026Vimarsha: Faithful ASR Evaluation for Indian Languages with Demographic Diversity, In-the-Wild Audio and Spelling Variations
  6. 2023Model Adaptation for ASR in low-resource Indian Languages
  7. 2026Factors affecting ASR performance: A study using state of the art ASR models in Indic Languages
  8. 2025Non-Intrusive Automatic Speech Recognition Refinement: A Survey