Using Deep Learning and Word Embedding to Detect Clickbait in Indonesian Headline News

Clickbait is a term used to describe provocative news article headlines designed to entice readers. Clickbait titles or headlines are frequently used to increase the number of clicks on articles or news stories so that the author or platform owner can profit. This will help potential readers avoid being provoked by a headline that contains clickbait. This study aims to detect clickbait in Indonesian article headlines by considering the effect of using Word2Vec and FastText as word embedding and analyzing the results of hyperparameter tuning performed on the model. The Bidirectional Long Short-Term Memory model will be applied to categorize news headlines in Indonesian article. The variation of the BiLSTM unit hyperparameter, dropout rate, learning rate, batch size, and epoch variation on the classification of clickbait in Indonesian news headlines will also be investigated in this study. The optimal hyperparameter value can be used in developing studies on models and future research. The best evaluation results are obtained when the model uses Word2Vec as word embedding with an accuracy of 0,892. For precision and recall obtained are 0,870 and 0,828.