ZYJ at SemEval-2021 Task 7: HaHackathon: Detecting and Rating Humor and Offense with ALBERT-Based Model

Proceedings of the 15th International Workshop on Semantic Evaluation (SemEval-2021), pages 1175–1178 Bangkok, Thailand (online), August 5–6, 2021. ©2021 Association for Computational Linguistics 1175 ZYJ at SemEval-2021 Task 7: HaHackathon: Detecting and Rating Humor and Offense with ALBERT-Based Model Yingjia Zhao Yunnan University / Yunnan, P.R. China zyj1309700118@gmail.com Xin Tao Yunnan University / Yunnan, P.R. China taoxinwy@126.com Abstract Although humorous language can bring joy to people, it is also easy to cause offense. There- fore, in order to effectively identify whether a sentence is humorous or offensive, the sys- tem needs to obtain abundant semantic infor- mation. This article introduces the submission of subtask 1 and subtask 2 that we participate in SemEval-2021 Task 7: HaHackathon: De- tecting and Rating Humor and Offense, we use a model based on ALBERT that uses ALBERT as the module for extracting text features. We modify the upper layer structure by adding spe- cific networks to better summarize the seman- tic information. Finally, our system achieves an F-Score of 0.9348 in subtask 1a, RMSE of 0.7214 in subtask 1b, F-Score of 0.4603 in sub- task 1c, and RMSE of 0.5204 in subtask 2. 1 Introduction A sense of humor is a positive psychological quality that can make people feel better about themselves, relieve stress, and gain a sense of connection with others. Humor can help bring about happiness experiences and optimism. Humor, by its nature, stirs up emotions. The speaker presents the audi- ence with an unexpected conflict. The audience feels nervous and anticipatory, and at the same time feels pleased and released. Humor is also a highly subjective phenomenon, with age, gender, and socioeconomic status is known to influence the perception of jokes. That’s why some things make people laugh and others don’t. When the brain lacks the cognitive resources to accurately understand the context in which a joke takes place, it generalizes it into an everyday behavior, and the benign offense becomes hostile aggression. Like most metaphorical languages, humor emphasizes multiple word meanings, cultural knowledge, and pragmatic capabilities, making it a challenging task to detect humor and offense in a sentence. SemEval-2021 Shared Task 7: HaHackathon: Detecting and Rating Humor and Offense (Meaney et al., 2021) shared task has two subtasks. Sub- task 1 emulates previous humor detection tasks in which all ratings were averaged to provide mean classification and rating scores. Subtask 1a: pre- dict if the text would be considered humorous (for an average user). This is a binary task. Subtask 1b: if the text is classed as humorous, predict how humorous it is (for an average user). The values vary between 0 and 5. Subtask 1c: if the text is classed as humorous, predict if the humor rating would be considered controversial, i.e. the variance of the rating between annotators is higher than the median. This is a binary task. Subtask 2 aims to predict how offensive a text would be (for an average user) with values between 0 and 5. This score was calculated regardless of whether the …

Content truncated to 500 words