Sentiment analysis became one of the core tasks in the field of Natural Language Processing especially with the rise of social media. Public opinion is important for many domains such as commerce, politics, sociology, psychology, or finance. As an important player in social media, Twitter is the
most frequently used microblogging platform for public opinion on any topic.
In recent years, sentiment analysis in Twitter turned into a recognized shared task challenge. In this thesis, we propose to enhance sentiment lexica with the linguistic notions negation and modality for this challenge. We test the interoperability between various sentiment lexica with each other and with negation and modality and add some Twitter-specific ad-hoc features. The performance of different combinations of these features is analyzed in comprehensive ablation experiments.
We participated in two challenges of the International Workshop on Semantic Evaluations (SemEval 2015). Our system performed robustly and reliably in the sentiment classification of tweets task, where it ranked 9th among 40 participants. However, it proved to be the state-of-the-art for measuring degree of sentiment of tweets with figurative language, where it ranked 1st among 35 systems.