Hinglish meeting notes with AI
· 4 min read
Hinglish meeting notes are hard because speakers switch between Hindi and English mid-sentence, and transcripts may be written in Devanagari, in Latin script, or both. Accurate results need a speech model trained on Indian languages and code-mixed speech, and an extractor that understands phrases like “kal tak” and “Friday tak”.
Why code-switching breaks ordinary tools
- Speech models tuned for one language mis-hear words from the other.
- The same word can appear in Devanagari or Latin script.
- Deadlines are relative: “kal” can mean yesterday or tomorrow, and weekday names are mixed with Hindi grammar.
What to look for
- Speech-to-text built for Indian languages with automatic language detection.
- Speaker labels, so each commitment is attributed to the right person.
- Date resolution that understands Hindi and Hinglish expressions relative to the meeting date.
- Owner matching that works with first names and nicknames.
Example
“Haan, main kal tak partners ko API docs bhej dunga” becomes the task “Send API docs to partners”, owner: the speaker, due: the day after the meeting, with the original sentence kept as evidence.
How Notemaker handles it
Notemaker transcribes with models built for Indian languages, lets the AI extract in English while keeping the original quote, and resolves relative dates in code from the quoted sentence so day-of-week mistakes do not slip through. Anything with an unclear owner or date is flagged for review.
Frequently asked questions
Can AI transcribe Hindi and English mixed together?
Yes, with speech models trained on code-mixed Indian speech. Accuracy depends on audio quality, accents and background noise, so review the output the first few times.
What does “kal tak” mean in a deadline?
“Kal” means both yesterday and tomorrow in Hindi, but in a commitment (“kal tak bhej dunga”) it almost always means by tomorrow. Good tools treat it that way and keep the quote so a human can check.