Live
Bangla NER classifier
State-of-the-art Bangla named entity recogniser serving 10 entity classes at 90.49% macro F1, deployed with FastAPI and TensorRT for the national Bangla corpus platform.
- PyTorch
- XLM-RoBERTa
- FastAPI
- TensorRT
- Node.js
The underlying model is an XLM-RoBERTa Large fine-tuned on the largest human-validated Bangla NER dataset (BanNERD). It recognises ten classes: Person, Organization, Geo-Political Entity, Location, Event, Number, Unit, Date & Time, Term & Title, and Miscellaneous.
Inference runs through TensorRT behind a FastAPI service, with a Node.js front end on the corpus platform.
The live demo is the corpus platform’s Tree Bank page, which runs this NER model alongside part-of-speech tagging, shallow parsing and dependency parsing.