SarcasmDet at SemEval-2022 Task 6: Detecting Sarcasm using Pre-trained Transformers in English and Arabic Languages

Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022), pages 1025 - 1030 July 14-15, 2022 ©2022 Association for Computational Linguistics SarcasmDet at SemEval-2022 Task 6: Detecting Sarcasm using Pre-trained Transformers in English and Arabic Languages Malak Abdullah and Dalya Faraj and Safa Swedat Jumana Khrais and Mahmoud Al-Ayyoub Computer Science, Jordan University of Science and Technology, Irbid, Jordan mabdullah@just.edu.jo dfalnore18, saswedat20, jdikhries162@cit.just.edu.jo maalshbool@just.edu.jo Abstract This paper presents solution systems for task 6 at SemEval2022, iSarcasmEval: Intended Sar- casm Detection In English and Arabic. The shared task 6 consists of three sub-task. We participated in subtask A for both languages, Arabic and English. The goal of subtask A is to predict if a tweet would be considered sarcastic or not. The proposed solution SarcasmDet has been developed using the state-of-the-art Ara- bic and English pre-trained models AraBERT, MARBERT, BERT, and RoBERTa with ensem- ble techniques. The paper describes the Sarcas- mDet architecture with the fine-tuning of the best hyperparameter that led to this superior system. Our model ranked seventh out of 32 teams in subtask A- Arabic with an f1-sarcastic of 0.4305 and Seventeen out of 42 teams with f1-sarcastic 0.3561. However, we built another model to score f-1 sarcastic with 0.43 in En- glish after the deadline. Both Models (Arabic and English scored 0.43 as f-1 sarcastic with ranking seventh). 1 Introduction In recent years, sarcasm has received remarkable research attention due to its importance and extraor- dinary impact on society, especially with the sig- nificant increase in its use in social media (Ghosh et al., 2020; Van Hee et al., 2018; Faraj and Abdul- lah, 2021). However, sarcasm is saying something, and what is meant is the opposite. Adding to the dif- ficulty of detecting sarcasm is that all the words in the sentence are positive, but the intended meaning is just the opposite (Channon et al., 2005). There- fore, the detection of sarcasm depends mainly on the general meaning of the sentence, which makes us move away from traditional methods and use artificial intelligence, especially natural language processing. Natural Language Processing (NLP) is a branch of artificial intelligence that gives machines the ability to read and understand the meanings of hu- mans (Nadkarni et al., 2011). As a result, ma- chines became able to understand the underlying meanings of words and not just rely on keywords. Moreover, it helps reveal forms of speech, such as emotion analysis, humor, ridicule, abuse, etc (Abujaber et al., 2021; Qarqaz et al., 2021). Task 6 at SemEval-2022, "iSarcasmEval: In- tended Sarcasm Detection In English and Arabic" (Abu Farha et al., 2022) suggested three main sub- tasks in both Arabic and English languages: Sub- Task A is a binary classification problem that deter- mines whether a tweet is sarcastic or not, SubTask B is a multi-label classification problem which de- termines the category of tweets sarcastic if it is found in English only, SubTask C is two sentences, one sarcastic and the other paraphrased to be non- sarcastic, define which one …

Content truncated to 500 words