Likewise, the Common Voice English dataset isn't great for ASR training outside India, either. There's a huge proportion of Indian speakers, and their data doesn't really help train ASR systems for non-Indian accents.
You and OP ("the indians labelers don't know how to do it correctly") seem to want the former, so it would be good to state that goal upfront.
A model is more useful if it approximates the speaker’s intent, not the listener’s interpretation. “Right” is complicated in language, but it’s hard to see how you’d use a model full of cross-cultural misunderstandings.
>English, due to its ‘lingua franca’ status, is an aspiration language for most Indians – for learning English is viewed as a ticket to economic prosperity and social status. Thus almost all private schools in India are English medium. Many public schools, due to political compulsions, have the state’s official languages as the primary school language. English is introduced as a second language from grade 5 onwards.
The paper claims:
>“All raters are native English speakers from India.”