How can voting mechanisms improve the robustness of individual toponym resolution approaches?

Abstract

This paper investigates the feasibility of implementing a general and robust toponym resolution approach by ensembling multiple existing approaches through voting mechanisms. Experiments were conducted to compare two voting ensembles with nine individual approaches based on seven public datasets. The results show that the voting ensembles can achieve consistent measures of Accuracy@161km and mean error, outperforming the individual approaches.