connotation_clashers at SemEval-2022 Task 6: The effect of sentiment analysis on sarcasm detection

Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022), pages 945 - 950 July 14-15, 2022 ©2022 Association for Computational Linguistics connotation_clashers at SemEval-2022 Task 6: The effect of sentiment analysis on sarcasm detection Patrick Hantsch and Nadav Chkroun University of Tübingen {patrick.hantsch, nadav.chkroun}@student.uni-tuebingen.de Abstract We investigated the influence of contradictory connotations of words or phrases occurring in sarcastic statements, causing those statements to convey the opposite of their literal meaning. Our approach was to perform a sentiment anal- ysis in order to capture potential opposite sen- timents within one sentence and use its results as additional information for a further classi- fier extracting general text features, testing this for a Convolutional Neural Network, as well as for a Support Vector Machine classifier, respec- tively. We found that a more complex and sophisti- cated implementation of the sentiment analysis than just classifying the sentences as positive or negative is necessary, since our implemen- tation showed a worse performance in both ap- proaches than the respective classifier without using any sentiment analysis. 1 Introduction According to Cambridge dictionary, sarcasm is de- fined as the use of remarks that mean the opposite of what they say, in order to hurt someone’s feel- ings or to criticize something in a humorous way. While detecting sarcasm can be very difficult even beyond the text level, it is a real challenge to detect sarcasm in textual data. When having conversa- tions in person, it is easier for people to detect sarcasm because they have additional information in form of the speaker’s emphasis of words, his facial expressions and body gestures. All of this information is not available in written language. Yet, detecting sarcasm from text becomes more and more important, since everyday life nowadays takes place on social media platforms to a large extent, unfortunately including things like hate speech in order to offend people. Sarcasm is often used as a tool for that (Frenda, 2018), so operators need ways to detect and remove offensive material expressed in a sarcastic way in large amounts of messages and posts released by millions of people every day. The shared task (described in Abu Farha et al., 2022) consisted of three sub tasks: • Sub task A: Determining if a given input text is sarcastic or not • Sub task B: Determining if a given input text belongs to a ironic speech category and if so, to which one exactly • Sub task C: Given two input texts, one be- ing sarcastic and the other one being its non- sarcastic rephrase, determining which is the sarcastic one We only participated in sub task A of the shared task. Our strategy was to additionally perform a sentiment analysis on the given text data, using the contradictory nature of a sarcastic statement, between the actual utterance and its true mean- ing/intent. For example in the sentence "I love how ill i became last night...", a sentiment analysis might recognize the clash of connotations between the positive connotated word "love" …

Content truncated to 500 words