Clickbait Detection in Indonesian News Sites Using Deep Learning

Abstract

Clickbait headlines are disruptive due to their deliberate exploitation of readers' curiosity, creating an information gap between what the reader knows and wants to know. These headlines often focus on sensationalized or misleading information, distorting the reader's perception. Therefore, addressing the issue of clickbait is crucial to maintaining the integrity of news and information dissemination and ensuring that readers are presented with accurate and valuable content that aligns with ethical journalistic standards.In this study, we analyzed clickbait in news headlines using a deep-learning approach. We tested the performance of two deep learning models, FastText+ Bidirectional Long Short-Term Memory (Bi-LSTM) and IndoBERT, to detect whether a news headline is clickbait or not. The results showed that both models can be effectively used as classification methods. Specifically, IndoBERT demonstrated superior accuracy compared to the FastText+Bi-LSTM approach, with an accuracy of 0.79. These findings suggest that IndoBERT is a more accurate and efficient solution for clickbait detection in news headlines. The findings of this study contribute to the ongoing efforts to address the issue of clickbait and its impact on news and information dissemination. By leveraging advanced deep learning techniques, this research provides valuable insights and tools for improving the quality and reliability of online news content, ultimately benefiting both readers and the broader field of journalism.