Fact · Languages and accents
In the LAHAJA study, recognition performance declined for speakers from North-East and South India, especially on named entities and specialized terms.
Checked against its source on October 17, 2026
In the source's own words
“the performance declines for speakers from North-East and South India, especially with content heavy in named entities and specialized terminology”
LAHAJA: A Robust Multi-accent Benchmark for Evaluating Hindi ASR Systems, Tahir Javed et al., arXiv preprint (AI4Bharat, IIT Madras), 2024, Abstract.
Open the original- What it measures
- The authors ran a fine-grained analysis by speaker region and by content type. Named entities are names of people, places and organizations.
- What it does not show
- The finding is about the test set and the models of 2024. It does not show that a given app will fail for a given speaker.