Sentiment Analysis in Social Media with a Semantic Web Based Approach: Application to the French Presidential Elections 2017

Abstract

Sentiment Analysis is one of the recent fields that attracts the attention of many researchers to contribute in its improvement and to get the fruits of its applications. This paper presents a novel framework that aims to empower the sentiment analysis task by combining an unsupervised approach that relies on a lexicon-based strategy and domain ontologies to identify the sentiment of the textual content, and a visual sentiment ontology to analyze the emotions expressed by images, this hybrid method ensure the accuracy of results on data retrieved from social networks to get insights about the reaction of public towards a specific topic. The framework was put into test to analyze the data flowing in social networks during French elections 2017 to rank the candidates and detect the regions where they are leading and the results obtained were promising.