Peer-reviewed research
Vistaar: Diverse Benchmarks and Training Sets for Indian Language ASR
Kaushal Santosh Bhogale et al., Interspeech 2023 (arXiv:2305.15386), 2023
Read the originalRead on October 17, 2026
- What it is
- Collects 59 benchmarks for Indian-language speech recognition, evaluates three public and two commercial systems on them, and introduces IndicWhisper, a Whisper model fine-tuned on Indian-language training data. The paper also reports a Hindi subset table comparing the systems.
- How much weight it carries
- Peer-reviewed at Interspeech 2023. The commercial systems were tested in 2023, so the comparison shows the gap at that time, not current accuracy. The authors built IndicWhisper themselves, so its lead is their own result and not an independent test.
Facts taken from this source
- On the Hindi subset of the Vistaar benchmark, IndicWhisper, which fine-tunes Whisper on Indian-language data, had an average word error rate of 13.6 percent across seven test sets.
“IndicWhisper 10.3 12.0 11.4 15.0 7.6 12.0 26.8 13.6” (Table 3)
- On the FLEURS subset of the Vistaar Hindi benchmark, IndicWhisper had a word error rate of 11.4 percent.
“IndicWhisper 10.3 12.0 11.4 15.0 7.6 12.0 26.8 13.6” (Table 3)
- In the Vistaar benchmark, IndicWhisper, a Whisper model fine-tuned on Indian-language data, had the lowest word error rate in 39 of the 59 benchmarks.
“IndicWhisper has the lowest WER in 39 out of the 59 benchmarks” (Abstract)
- The Vistaar benchmark collects publicly available Indian-language test sets across 12 languages, giving 59 computed word error rate values.
“comprises of publicly available benchmarks across 12 languages, leading to 59 computed WER values across benchmarks and languages” (Introduction)