Intuition dictates that figurative language and especially metaphorical expressions should convey sentiment. It is the aim of this work to validate this intuition by showing that figurative language (metaphors) appearing in a sentence drive the polarity of that sentence. Towards this target, the current thesis proposes methods for the sentiment analysis of sentences where figurative language plays a dominant role. In particular we explore differentMachine Learning methods, as well as the potential of a Machine Learning method to function complementarily with a rule-based one. The proposed method applies Word Sense Disambiguation in order to assign polarity to word senses rather than words. Sentence polarity is determined using the individual polarities for metaphorical senses as well as other contextual information.
In the course of our experiments we found that many of the cases for which our method yielded erroneous evaluations bear mild figurativeness. This led to a combined method in which these cases were processed by a rulebased system which we support it is better for cases close to literal language. These two systems, acting complementarily, can address limitations of machine learning methods, especially in cases where gurative language is not that strong. For each Sentiment Analysis method, we present experimental results on benchmark corpora that indicate the potential of the proposed methods. Finally, experimental results provide supportive evidence that the methods presented are also well suited for corpora consisting of literal and figurative language sentences.