SarcasmDet at SemEval-2021 Task 7: Detect Humor and Offensive based on Demographic Factors using RoBERTa Pre-trained Model

Proceedings of the 15th International Workshop on Semantic Evaluation (SemEval-2021), pages 527–533 Bangkok, Thailand (online), August 5–6, 2021. ©2021 Association for Computational Linguistics 527 SarcasmDet at SemEval-2021 Task 7: Detect Humor and Offensive based on Demographic Factors using RoBERTa Pre-trained Model Dalya Faraj Department of Computer Science Jordan University of Science and Technology Irbid, Jordan dfalnore18@cit.just.edu.jo Malak Abdullah Department of Computer Science Jordan University of Science and Technology Irbid, Jordan mabdullah@just.edu.jo Abstract This paper presents one of the top winning so- lution systems for task 7 at SemEval2021, ”Ha- Hackathon: Detecting and Rating Humor and Offense”. The shared task 7 consists of two parts, task-1 with three sub-tasks 1a,1b, and 1c, and task-2. The goal of task-1 is to predict if the text would be considered humorous or not, then if it is yes, predict how humorous it is and whether the humor rating would be perceived as controversial. The goal of task-2 is to pre- dict how the text is considered offensive for users in general. The proposed solution, Sar- casmDet, has been developed using RoBERTa pre-trained model with ensemble techniques. The paper describes the submitted system’s ar- chitecture with the experiments and the hyper- parameter fine-tuning that led to this robust system. Our model ranked third and fourth places out of 50 teams in tasks 1c and 1a with F1-Score of 0.6270 and 0.9675, respectively. At the same time, the model ranked one of the top 10 models in task 1b and task 2 with RMSE scores of 0.5446 and 0.4469, respec- tively. 1 Introduction In our daily life, the obstacles and difficulties in dealing with sarcasm, bullying, or even abuse of all kinds and ways are increasing day by day (Sheehan et al., 1999; Cleary et al., 2009; Tucker and Maun- der, 2015; van Verseveld et al., 2021). Technically, sarcasm and bullying are among the most complex and challenging topics that major companies and institutes seek to address. Artificial intelligence and text processing techniques are the most po- tent current methods for detecting these problems within texts and images. Sarcasm and abuse are associated with attacking a specific person or group of people either through an unintended joke or, in many cases, by directly affecting the target’s psy- che. Irony and offensiveness are characterized by their vocabularies that are peppered with humor to conceal the opposite (Lee and Katz, 1998). Task 7 at SemEval-2021, ”HaHackathon: De- tecting and Rating Humor and Offense”, provides two main tasks: task-1 with three sub-tasks (1a,1b, 1c) and task-2. The goal of task-1 is to predict if the text would be considered humorous or not, and if it is yes, then expect how funny it is and whether the humor rating would be perceived as controversial. The goal of task 2 is to predict how the text is con- sidered offensive for users in general. Our solution, SarcasmDet, has been ranked among the top four teams in two sub-tasks. The proposed approach uses the provided dataset, which contains 10K of row text …

Content truncated to 500 words