The challenge
People, places and other named entities are difficult for generic NLP models to recognise reliably in low-resource language contexts.
Approach
- 01Develop and document labelled named-entity data.
- 02Train and evaluate a sequence labelling model.
- 03Describe dataset limitations and evaluate at entity level.
Documented outcome
The repository reports 0.7034 entity-level F1 on its documented test set.
Interpretation & limitations
The score applies to the documented dataset and label policy; further validation is needed for different dialects and use cases.