Relevance as an enhancer of votes on Twitter

Abstract

The concept of the influence of Katz and Lazarfeld given in the last century has evolved thanks to the appearance of Social Networks and especially Twitter. Because this microblogging has allowed candidates for any election process to be closer to their electors and also allows an analysis of the contents of the messages to determine their polarity. The relevance of the messages that measure the level of influence that can be had in the voters, incorporated into the traditional analysis of the Social Networks allow to have a greater degree of precision in the electoral predictions that are made using natural language processing, NLP. We have introduced in the methodology that we propose a mechanism to enhance the votes of those messages that have a greater relevance and turn them into votes in order to improve the predictability of the electoral results. The proposed methodology was applied in the election for President of the Republic of Ecuador that was held on February 19, 2017, obtaining a Mean Average Error, MAE = 1.4 that demonstrates the relevance of incorporating the variable Relevance as an enhancer of votes.