English and Arabic Sarcasm Detection in Tweets (2022)

Team

• Amy Tzu-Yu Chen @amy17519
• David Roesler @droesler
• Diana Baumgartner-Zhang @Diana-BZ
• Juliana McCausland @ju-mc

  Description

  Group project for University of Washington LING573 NLP Systems and Applications Spring 2022. We build English and Arabic sarcasm detection systems using the Twitter sarcasm dataset of SemEval 2022 Task 6, iSarcasmEval (Oprea and Magdy, 2020). Our system consisting of a fine-tuned BERTweet with threshold movement and data augmentation outperforms the top scoring submission for the English portion of the task by 4 points F1.

• Our paper and slides for the project.
• Github repository link.

  Primary task results

System results for positive (sarcastic) class.