Sentiment analysis for Turkish Language
Hakan Çelik · 2013
Sentiment analysis a.k.a. opinion mining is an application of natural language processing to extract subjective information from given texts. Consumer comments, evaluation of films, stock exchange predictions and political researches are some examples of subjective declarations that sentiment analysis involves. Subjective data have hugely increased parallel to web 2.0 and social media growth in recent years. People generate petabytes of data every day. Processing of this amount of raw data became more and more important for both companies and individuals. Many sentiment analysis studies conducted in the past are focused on English Language only. Therefore, for Turkish Language, sentiment analysis is a blue ocean yet. There is a single study focused on only one input domain and tested with only one machine learning algorithm. Within the focus of this thesis new datasets were constructed in various domains, a Turkish specific input preparation algorithm was introduced, and different machine learning algorithms were applied over prepared data. ~85% accuracy was achieved with Naive Bayes Classifier.