hub at SemEval-2021 Task 7: Fusion of ALBERT and Word Frequency Information Detecting and Rating Humor and Offense

Proceedings of the 15th International Workshop on Semantic Evaluation (SemEval-2021), pages 1141–1145 Bangkok, Thailand (online), August 5–6, 2021. ©2021 Association for Computational Linguistics 1141 hub at SemEval-2021 Task 7: Fusion of ALBERT and Word Frequency Information Detecting and Rating Humor and Offense Bo Huang School of Information Science and Engineering Yunnan University, Yunnan, P.R. China hublucashb@gmail.com Yang Bai School of Information Science and Engineering Yunnan University, Yunnan, P.R. China baiyang.top@gmail.com Abstract This paper introduces the system description of the hub team, which explains the related work and experimental results of our team’s participation in SemEval 2021 Task 7: Ha- Hackathon: Detecting and Rating Humor and Offense. We successfully submitted the test set prediction results of the two subtasks in the task. The goal of the task is to perform hu- mor detection, grade evaluation, and offensive evaluation on each English text data in the data set. Tasks can be divided into two types of sub- tasks. One is a text classification task, and the other is a text regression task. What we need to do is to use our method to detect the humor and offensive information of the sentence as accurately as possible. The methods used in the results submitted by our team are mainly composed of ALBERT, CNN, and Tf-Idf algo- rithms. The result evaluation indicators sub- mitted by the classification task are F1 score and Accuracy. The result evaluation index of the regression task submission is the RMSE. The final scores of the prediction results of the two subtask test sets submitted by our team are task1a 0.921 (F1), task1a 0.9364 (Accu- racy), task1b 0.6288 (RMSE), task1c 0.5333 (F1), task1c 0.0.5591 (Accuracy), and task2 0.5027 (RMSE) respectively. 1 Introduction and Background Perceiving humor has always been a unique ability of human beings. So what is the use of humor? The research results of Martin and Kuiper on hu- mor show us the influence of humor on a person’s physical and mental health (Martin, 2004; Kuiper et al., 2004). In recent years, the use of automated methods to detect humorous information in text has attracted widespread attention (Barbieri and Saggion, 2014; Reyes et al., 2012). SemEval 2021 Task 7: Ha-Hackathon: Detect- ing and Rating Humor and Offense’s task goal is to use automated techniques and methods to automat- ically detect humor and grade in the text. Besides, this task also needs to evaluate the offensive level of the text data. The task is divided into two parts, one part is the detection and evaluation related to humor. There are three subtasks in this part of the task. It involves text classification and regression. The other part is to assess the offensive level of the text data. This is a separate text regression task. The purpose is to predict how offensive the text will be to ordinary users. This task is an interest- ing challenge for the machine. Humor is a very subjective emotion. People with different cultural backgrounds and life experiences have different feelings about the same sentence. …

Content truncated to 500 words