Clemenciah/AfriSQuAD

AfriSQuAD is a dataset created in the paper:Evaluating the Robustness of Machine Reading Comprehension Models to Low Resource Entity Renaming accepted at AfricaNLP'23 # Evaluating the Robustness of Machine Reading Comprehension Models to Low Resource Entity Renaming This repository hosts a project on evaluating the generalization capability of MRC models and geo representation of MRC datasets which was accepted at the AfricaNLP workshop 2023 co-located with ICLR in Kigali Rwanda. We investigate: 1. How diverse are the named entities in the SQuAD dataset in terms of geolocation? 2. How do QA models perform when an entity is swapped from the original dataset with a low-resource named entity? The data folder hosts the swapped dev set of SQuAD and named entity gazeteers generated with African-origin entities