Parliamentary election’s predictions using social media content

Antoni Sobkowicz, Marek Kozłowski

2019 W: PP-RAI'2019 : Polskie Porozumienie na rzecz Rozwoju Sztucznej Inteligencji : Conference Proceedings / Paweł Zyblewski, Paweł Ksieniewicz; Wrocław: Politechnika Wrocławska, s. 184-188

Political news are a major part of many information websites and are often emotionally commented on by a huge number of users. Using such data (counted in millions of comments) we searched for an automated way to find coherent groups of users that produce content that is likely to evoke negative emotions towards political parties. Combining time patterns with machine learning methods we build a toolkit to infer an opinion poll using web public data. We proved that such methodology brings very similar results in comparison to the classical approaches used by major opinion polling institutes.