TOBB ETU at CheckThat! 2023: Utilizing ChatGPT to Detect Subjective Statements and Political Bias

Abstract

Information has been referred to as the “oil” of the 21st century, emphasizing its immense importance. However, it also has the potential to pose significant risks and hazards if it is not correct. Hence, it is imperative to reduce the spread of misinformation. In this paper, we present our participation in Task 2 (i.e., detecting subjective tasks) and Task 3A (i.e., detecting political bias in news articles) of CLEF CheckThat! 2023 which focuses on reducing the spread of misinformation. We propose utilizing ChatGPT for these classification tasks and explore zero-shot and few-shot classification using ChatGPT. While the performance of our approach varies across different languages in Task 2, we are ranked 3rd on the German dataset with 0.71 macro F1 score. In Task 3A, we are ranked 2nd with 0.646 Mean Absolute Error (MAE). © 2023 Copyright for this paper by its authors.