English–Indian language translation helper
Build and evaluate a translation helper for an Indian language, tuned for one field such as education or health.
The brief
Government and health information is mostly in English. Build a translation helper for English and one Indian language, fine-tune it for one field (for example health advice), and evaluate it with both automatic scores and native speakers. Good looks like: a measured gain from fine-tuning, a human evaluation, and a glossary that keeps key terms consistent.
What you will build
- Choose the language pair and field
- Prepare parallel sentences and a test set
- Fine-tune a pre-trained translation model
- Evaluate with BLEU/chrF and native speakers
- Add a glossary for key terms
- Build a demo and write the report
Free, open data to build it with
- Samanantar parallel corpus ↗ AI4Bharat (Hugging Face)
Check each dataset's licence before you publish your work. Your mentor gives your team its own version of the brief (data, scope or region), so every team's project is its own.

