In this paper, we describe our system submitted to SemEval 2021 Task 7: HaHackathon: Detecting and Rating Humor and Offense. The task aims at predicting whether the given text is humorous, the average humor rating given by the annotators, and whether
the humor rating is controversial. In addition, the task also involves predicting how offensive the text is. Our approach adopts the DeBERTa architecture with disentangled attention mechanism, where the attention scores between words are calculated based on their content vectors and relative position vectors. We also took advantage of the pre-trained language models and fine-tuned the DeBERTa model on all the four subtasks. We experimented with several BERT-like structures and found that the large DeBERTa model generally performs better. During the evaluation phase, our system achieved an F-score of 0.9480 on subtask 1a, an RMSE of 0.5510 on subtask 1b, an F-score of 0.4764 on subtask 1c, and an RMSE of 0.4230 on subtask 2a (rank 3 on the leaderboard).
Humor detection and rating poses interesting linguistic challenges to NLP; it is highly subjective depending on the perceptions of a joke and the context in which it is used. This paper utilizes and compares transformers models; BERT base and Large,
BERTweet, RoBERTa base and Large, and RoBERTa base irony, for detecting and rating humor and offense. The proposed models, where given a text in cased and uncased type obtained from SemEval-2021 Task7: HaHackathon: Linking Humor and Offense Across Different Age Groups. The highest scored model for the first subtask: Humor Detection, is BERTweet base cased model with 0.9540 F1-score, for the second subtask: Average Humor Rating Score, it is BERT Large cased with the minimum RMSE of 0.5555, for the fourth subtask: Average Offensiveness Rating Score, it is BERTweet base cased model with minimum RMSE of 0.4822.
