Dynamic profiles using sentiment analysis and twitter data for voting advice applications

Nowadays, political campaigns combine traditional media channels with social media platforms, opening new and promising possibilities for parties and candidates looking for better political strategies and visibility. Voting advice applications (VAAs) recommend parties and candidates that are close to a citizen's political preferences and require the constrution of candidate and party profiles. Profile generation is an essential task in the development of VAAs and requires two steps: an unbiased design of political questionnaires and the collection of all candidates' answers. This paper presents an extension of a VAA, implemented in within the project Participa Inteligente (PI), a social-network platform designed for the 2017 Ecuadorian national elections. This work concentrates on the implementation of dynamic candidate profiling using Twitter data and sentiment analysis as an additional element to the static profile generation of VAAs. The implementation of a dynamic element for VAAs could help mitigate the effect of biased recommendations given during the construction of candidate and party profiles. At the end of this work, the dynamic profile is compared with the classic static elements developed within the PI project. The results show the level of similarities and differences between each of the elements in profile generation. This work provides an ideal basis for future research in the area of VAAs and their interfaces. Additionally, it opens up a broader spectrum of applications for policymakers including decision-making and collaborative working environments toward e-empowerment.