Bilkent University
Department of Computer Engineering


Predicting Sentiment Strength of Turkish Tweets


Esra Akbaş
MSc Student Student
Computer Engineering Department
Bilkent University

Recently, people share their views and recommendations through the web such as public social media. As the type and amount of venues for sharing opinions increase, analyzing sentiment on textual resources has become an essential data mining goal. Sentiment classification aims to identify and categorize the polarity of sentiment in text. The polarity is predicted on either a binary; positive, negative or multivariant scale as the strength of sentiment expressed. Text often contains a mix of positive and negative sentiment, hence it is often necessary to detect both simultaneously. In my thesis, I focus on Turkish texts in Twitter that contains informal short messages. After constructing and preprocessing a Turkish data repository, and generating a list of Turkish words with their sentiment strengths, I design a new type of feature vectors that aim to detect positive and negative sentiment. I utilize several machine learning methods to generate classifiers and test their performance.


DATE: 28 November, 2011, Monday @ 16:15