The Story Behind BENI
It started with a simple question: Why is all economic NLP research about English?
84% of sentiment-based forecasting focuses on developed economies. Bangla, the 7th most spoken language in the world? Almost nothing.
The Dataset
BENI v1.0 is a Bangla Economic Narrative Index dataset. 620K+ articles across 10 languages from the Global South. Published on HuggingFace and arXiv.
What I Learned
- Low-resource NLP is hard because the data doesn't exist.
- Economic narratives matter — they shape policy, markets, perceptions.
- Building the dataset IS the research.
Read the paper: arXiv:2606.10225