Regresar

Analyzing and integrating dynamic profiles on voting advice applications

Abstract:

Voting advice applications (VAA's) are interactive online tools that have become extremely popular in electoral campaigns. They are developed to assist voters by enhancing the basis on which they decide how to vote. Typical VAA's do this by matching users' policy-preferences with the positions of parties and/or candidates. Nowadays, VAA's are able to attract huge numbers of respondents and to provide a significantly rich source of mass public opinion data. However, not all parties or candidates provide their positions regarding the statements revealed in the VAA platform in order to derive recommendations for users. The majority of candidate profiles are constructed using expert analysis, extracting the position of parties and/or candidates from different data sources, such as interviews, speeches, and discussions, among others. This type of profile generation is time-consuming and sometimes arguable, which may lead to mismatching or biased results. Therefore, it is important to find a dynamic approach for building candidate profiles. In this paper, the authors propose a so-called VAA 2.0, which generates dynamic profiles of politicians by extracting their attitude toward policy-issue statements from their official Twitter accounts. The VAA 2.0 works with the Twitter API, a directional model for matching algorithm, an affective norm for English words (ANEW) library, the cumulative distribution function, and the probability density function for sentiment analysis.