This repository presents the results of the research conducted in June 2021 - December 2021.
In the process of statement selection for voting advice applications (VAA), political scientists spend a considerable amount of time analyzing the domestic and foreign policies of a given country. While there is a large amount of government open data available to the public, analyzing these data manually is impractical. In order to facilitate such time-consuming and labor-intensive work, the authors propose a system to assist VAA designers and developers in re-designing statements. Using advanced language modeling and text summarization techniques on the VAA data and a politics-related corpus, the system produces suggestions that would be applicable for revising the original VAA statements. Experiments conducted on Taiwanese VAA and e-petition data show that the system can generate valuable suggestions for VAA designers and could serve as an additional tool for VAA development.
All the suggestions the system produced and the labeled topics with their descriptions are available here.