Detecting Sarcasm in Communication on Twitter

Sarcasm is a complex multimodal phenomenon that can be difficult to detect even in face-to-face communication, meaning it may be even harder to recognize in communication on social media, considering the absence of intonational and gestural cues. This thesis deals with the ways in which sarcasm is delivered and recognized on Twitter. A manual qualitative analysis of 31 tweets showed that Twitter users try to replicate the same principles that are used in oral communication, firstly, by using the same fixed lexical expressions, interjections and intensifiers, and secondly, by substituting intonational cues with punctuation, word and phrase capitalization and letter duplication, and gestural cues with emojis, images and GIFs. Although the presence of most of these cues by itself does not necessarily indicate sarcastic intent, the results showed that most Twitter users combine multiple multimodal cues, which makes sarcasm detection easier