Abstract:
Determining the feelings of a user on a topic is called sentiment analysis. Senti ment analysis is done so as not just to get a sort of feedback data for decision making but also to check the emotions, characteristics and influence of situation actions, news or events on users or specific data with respect to users. Sentiment analysis of a sen tence is the effort of understanding whether the sentence contains positive, negative or neutral meaning and is used to gather overall attitude toward a product, person or event etc. In this thesis we have studied the effectiveness of applying machine learn ing and statistical natural language processing techniques collaboratively to Twitter messages based sentiment classification problem. This thesis covers the work on Turk ish tweet sentiment analysis, sector based sentiment analysis, English tweet sentiment analysis, and political opinion prediction. We introduce sector based sentiment analysis framework by applying machine learning and statistical natural language techniques collaboratively. We apply our framework to finance, telecom, retail and sport sectors. In English sentiment analysis, this thesis covers our solution for the SemEval Twitter Sentiment Analysis task. Consecutively, in English Twitter analysis our work also ad dresses the task of political orientation classification based on Twitter data. We have used various machine learning algorithms that can predict and automatically classify whether a tweet belongs to a republican or a democrat. We have used a Twitter dataset which consists of democrat and republican voters tweets. This work covers political opinion/sentiment tendency estimation based on Twitter messages of voters.
