Twitter opinion about leaders

Abstract

Mutual respect between leaders and followers is a key prerequisite to success. The opinion of followers in challenging this leadership is just as great as it has been portrayed by the uprisings in North Africa and the Middle East tagged as the “Twitter or Social media revolution”. The sudden eruption of activities in the area of opinion mining, which deals with the computational analysis of opinion, sentiment, and subjectivity in text, has thus occurred as a means of responding directly to the surge of interest that deals with opinions and use of information technologies to seek out and understand the opinions of others. This study focused on identifying a set of suitable features and an appropriate classifier that can be used for detecting and classification of opinions about leaders in tweets. Words, unigram, bigram and negation features were used alongside Naïve Bayes (NB) and Support Vector Machine (SVM) learning algorithms. The results show that using NB with unigrams can indicate opinions about leaders of up to 91.41% accuracy and can therefore be used to suggest ways to improve a leader’s reputation as well as predicting potential candidates in political election.