Our dataset consisted of 101 theaters, 100 museums, 100 towers, 100 universities, 204 companies, 123 sport teams and 125 politicians. The link for each class gives the names and the ground truth, one name per row.
Nima Miryeganeh has implemented some of the algorithms in the paper (excluding toponym resolution) and has made it public here.
Relevant paper (to be cited if the dataset is used)
Jiangwei Yu, Davood Rafiei: (2016). Geotagging named entities in news and online documents, Proceedings of the ACM Conference on Information and Knowledge Management (CIKM), Indianapolis, 2016.