Automatic Sarcasm Detection in Twitter Messages
MetadataShow full item record
In the past decade, social media like Twitter have become popular anda part of everyday life for many people. Opinion mining of the thoughtsand opinions they share can be of interest to, e.g., companies and organizations.The sentiment of a text can be drastically altered when figurativelanguage such as sarcasm is used. This thesis presents a system for automaticsarcasm detection in Twitter messages.To get a better understanding of the field, state-of-the-art systems fordetecting sarcasm in Twitter messages are explored. Many such systemsalready exist, and a common theme among them is the use of automaticallyannotated data for both training and testing. In addition to presenting asystem for detecting sarcasm, this thesis also looks into the use of manuallyannotated data for testing. To this end, a dataset of tweets manuallyannotated with respect to the presence of sarcasm was built. The resultwas very similar to that of a previously made set, and both of them showedconsiderable deviation from automatic annotation. This implies that usingautomatically annotated data for the task of sarcasm detection in tweetsis a mediocre approximation.Experiments with both of the manually annotated datasets also gavevery similar results, showing that they are well annotated and reasonablyrepresentative for sarcasm detection in tweets.