Factors affecting ASR performance: A study using state of the art ASR models in Indic Languages
On the map: Speech and audio language models → Multilingual and low-resource ASR
Abstract (arXiv)
ASR performance varies across languages, speakers, and recording conditions, yet systematic analysis for Indic languages remain limited. We present a large-scale study of decoded outputs from multiple open-source ASR models evaluated on diverse Indian speech datasets in zero-shot settings. We analyze linguistic, speaker-level, and acoustic factors across Hindi, Bengali, Kannada, Telugu, and Marathi. We examine correlations between WER and speaker traits such as average word length, speaking rate, and utterance duration across multiple model dataset pairs. For Hindi, we further analyze audio factors including telephone codecs, bit depth, resampling, and background noise. Results reveal both cross lingual patterns and language-specific sensitivities, showing how speaker behavior and signal processing choices affect ASR robustness in real world Indic scenarios.
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 12 of its 27 references that are in the map. First posted on arXiv 8 Jun 2026.
Builds on
12 of 27 references.
Closest ancestors · 3
Same direction, newest first
- 2022IndicSUPERB: A Speech Processing Universal Performance Benchmark for Indian languages24 Aug
- 2022FLEURS: Few-shot Learning Evaluation of Universal Representations of Speech25 May
- 2021CLSRIL-23: Cross Lingual Speech Representations for Indic Languages15 Jul
Other references · 9
- 2017Attention Is All You Needcited by 61,180
- 2022Robust Speech Recognition via Large-Scale Weak Supervisioncited by 4,661
- 2019Common Voice: A Massively-Multilingual Speech Corpuscited by 1,596
- 2022data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Languagecited by 774
- 2020Massively Multilingual ASR: 50 Languages, 1 Model, 1 Billion Parameterscited by 117
- 2021VoiceFixer: Toward General Speech Restoration with Neural Vocodercited by 57
- 2023AudioSR: Versatile Audio Super-resolution at Scalecited by 45
- 2026VAANI: Capturing the language landscape for an inclusive digital Indiacited by 14
- 2022data2vec-aqc: Search for the right Teaching Assistant in the Teacher-Student training setupcited by 6
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026A study on the impact of region specific data on the performance of Indic ASR
- 2021Towards Building ASR Systems for the Next Billion Users
- 2025Effects of Speaker Count, Duration, and Accent Diversity on Zero-Shot Accent Robustness in Low-Resource ASR
- 2026Voice of India: A Large-Scale Benchmark for Real-World Speech Recognition in India
- 2024Advocating Character Error Rate for Multilingual ASR Evaluation
- 2026Vimarsha: Faithful ASR Evaluation for Indian Languages with Demographic Diversity, In-the-Wild Audio and Spelling Variations
- 2026Responsible ASR: Overcoming Challenges of Foundational Models in Narrow-Band and Low-Resource Settings
- 2025The ML-SUPERB 2.0 Challenge: Towards Inclusive ASR Benchmarking for All Language Varieties