Sentiment analysis for Indian languages

The brief

A brand gets reviews in Hindi, Tamil, Bengali and other languages but only analyses English ones. Build a sentiment classifier for Indian-language reviews and compare a multilingual model with translation-then-English approaches. Good looks like: accuracy reported per language, a fair comparison of approaches, and examples of tricky cases.

What you will build

1. Explore the reviews per language
2. Build a baseline with translation to English
3. Fine-tune a multilingual model
4. Compare results per language
5. Discuss errors with examples

Free, open data to build it with

• IndicSentiment ↗ 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.