| Abstract |
With the constant social, economic, and political changes witnessed in the world, numerous events occur in cities, and lead to important reactions from the public, including riots, civil disorder, and violent actions. During those events, situational awareness is crucial to gain a good understanding of the events and their impact on public opinion, which is typically difficult to measure. Opinion leaders are influential people who are able to shape the thoughts of others in the society, through eloquent and inspirational opinions and posts. Analyzing the sentiments of opinion leaders can give a strong indication about the sentiment in the street, due to the high influence that those leaders have on their followers. The purpose of this work is to build an opinion leaders' sentiment monitoring framework that would serve as decision support tool for government officials. Our framework leverages our previously proposed opinion leader identification algorithm, along with text mining, text classification, and sentiment annotations to extract sentiment intelligence from opinion leaders' posts, for effective analysis of public opinion about ongoing events. The proposed framework was implemented and tested using datasets collected from 27,000 Twitter accounts over a 15 months' span. Opinion leaders were identified in five domains (economics, politics, health, sports, and education) based on their competency and popularity. Furthermore, a linear Support Vector Machine (SVM) classifier along with sentiment annotations were used to perform sentiment analysis on the tweets posted by 43 opinion leaders in relation to a major political event. The results obtained are very promising and indicate the potential of leveraging high impact social media content to gain insights about public opinion. © 2019 University of Split, FESB. |
