Pattern Based Features for Sarcasm Detection on Reddit

Abstract

Sarcasm is a complex social mechanism which reverses the polarity of a sentence. In the past, research has been done extensively on twitter due to the large amount of data that can be extracted. In this thesis, we explore sarcasm detection on reddit, a platform which is different from twitter. The results indicate a 59% accuracy of the classifier and 65,9% accuracy when filtering the comments on subreddit, indicating above chance classification but also that there is much to improve.