[3M1-GS-10] AI application

[3M1-GS-10-03] Detecting sarcastic expressions for extracting sarcasm in a text

〇Reon Hata¹, Yutaka Morino¹, Mitsunori Matsusita¹ (1. Kansai University)

Keywords:sarcasm detection, sarcasm situation, BERT, Twitter

Sarcasm is a linguistic phenomenon in which a target’s faults and weaknesses are criticized indirectly and is characterized by the use of positive expressions to convey negative meanings. For this reason, an expression’s superficial meaning and intention do not match in sarcastic sentences. When using a computer to determine whether an affirmative expression appears in a sentence is sarcasm, it is necessary to extract the three elements of sarcasm together: the target of the sarcasm, the situation that is the premise of the sarcasm, and the sarcastic expression. Although previous studies have attempted to identify sarcastic sentences by extracting the target of sarcasm, there are many sarcastic sentences in which the target of sarcasm is not explicitly stated. Therefore, in this study, we create a classifier for detecting sarcastic situations from sentences expressing situations using BERT and attempt to extract sarcastic sentences in which the target of the sarcasm is not explicitly stated.