IronyMagnet at SemEval-2018 Task 3: A Siamese network for Irony detection in Social media

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 570–575 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics IronyMagnet at SemEval-2018 Task 3: A Siamese Network for Irony Detection in Social Media Aniruddha Ghosh University College Dublin Dublin, Ireland aniruddha.ghosh@ucdconnect.ie Tony Veale University College Dublin Dublin, Ireland tony.veale@ucd.ie Abstract This paper describes our system, entitled IronyMagnet, for the 3rd Task of the SemEval 2018 workshop, “Irony Detection in English Tweets”. In Task 1, irony classification task has been considered as a binary classification task. Now for the first time, finer categories of irony are considered as part of a shared task. In task 2, three types of irony are considered; “Irony by contrast” - ironic instances where evaluative expression portrays inverse polarity (positive, negative) of the literal proposition; “Situational irony” - ironic instances where output of a situation do not comply with its expectation; “Other verbal irony” - instances where ironic intent does not rely on polarity contrast or unexpected outcome. We proposed a Siamese neural network for irony detection, which is consisted of two subnetworks, each containing a long short term memory layer (LSTM) and an embedding layer initialized with vectors from Glove word embedding1. The system achieved a f-score of 0.72, and 0.50 in task 1, and task 2 respectively. 1 Introduction Irony is one of the most prominent and pervasive figures of speech in human communication, dat- ing back to ancient religious texts to modern mi- crotexts. According to literary scholars (Grice, 1978; Lakoff, 1993), irony has been defined as a trope where the speaker intends to communicate a contradictory situation or the opposite meaning of what is literally said. It adopts a subtle tech- nique where incongruity is used to suggest a dis- tinction between reality and expectation in order to produce a humorous or emphatic effect on the listener. Irony poses an important challenge not only from a linguistic perspective but also from a cognitive one. Even without a solid understand- ing of irony, one can still recognize and produce 1https://nlp.stanford.edu/projects/glove/ ironic utterances from as early as childhood (Har- ris and Pexman, 2003). Such capabilities are of- ten associated with one’s ability to correctly in- fer others’ communicative intentions and perspec- tives towards a given situation. Psychological the- ories, such as “echoic reminder theory” (Kreuz and Glucksberg, 1989), “allusion pretense theory” (Kumon-Nakamura et al., 1995), and “implicit dis- play theory” (Utsumi, 2000), confirm that cues for understanding ironic intent are not restricted to language. Ironic intent involves several other aspects including, but not limited to, the context of an utterance, the world’s perception and famil- iarity between the listener and the speaker, and psychological dimensions. Thus, as a purely text classification task, the irony detection task poses a significant challenge for computational linguists. Computational approaches focus on identifying the subtle incongruity between different parts of the text. Often, an ironic statement starts with an overtly positive attitude (“Yay I love”) and ends up in an disappointment (“working on my birth- …

Content truncated to 500 words