Path: Bharat language & voice
Respect dialect and register
Judge a dialect against the people who speak it, and separate a real mistake from a difference from the standard.
- Starting level
- Some basics
- Time
- 18 minutes
- Public lesson guide
- EN + हिंदी
You will learn to
Separate a dialect feature from a speaker mistake, and score against the variety the guideline names.
Why this matters
A dataset that quietly corrects dialects teaches a model that crores of speakers speak their own language wrongly.
Three things to remember
- A dialect feature is not a mistake to be standardised.
- Judge the language against the variety the guideline names.
- Record the feature you noticed instead of editing it away.
See one example
A Bundeli clip uses 'हम' for one person; the guideline asks for the speaker's own variety.
Try the judgement yourself
Both answers respond to the task above. One of them does the job better. Which, and why?
Make something yourself
Take a sentence in a language variety you know well that a standard-language teacher would mark wrong. Explain what it does, who says it, and why it is not an error.
Read further
The source is optional. The complete teaching and practice are available inside AI2Bharat.
Read the primary or official source: AI4Bharat