Sentiment analysis of social network posts in Slovak language

Abstract

In this paper we tackle the issue of sentiment analysis of social network posts in a not well targeted language - Slovak. There is a significant lack of research in this area for minor languages, as they often introduce additional language-specific issues for text processing. In case of Slovak, common issues are high flection, complex morphology and syntax. User-generated content of social networks introduces additional challenges (variability of diacritics, inconsistent style, high error rate) that make the task even harder. In this paper, we propose a method for sentiment analysis of social network posts on Facebook. The proposed method is based on machine learning and incorporates multilevel text pre-processing aiming to deal with specifics of user-generated social content. The evaluation in a real-word setting employing data from Facebook pages of multiple well-known companies shows accuracy of our method comparable with approaches for major world languages.