Clickbait Detection Using Bi-LSTM Model

Abstract

Clickbait describes the information that aims to catch readers' attention and influence users to click a link or interact with the content, such as headlines or thumbnails. To encourage viewers and attract them to go through and read the post or article, clickbait frequently uses exaggerated or flashy language, provocative pictures, or missing information. Although clickbait can be useful for increasing website traffic, it can also be deceptive or manipulative and might not live up to the claims made in the title or thumbnail. Because of this, many individuals consider clickbait to be annoying or a type of dishonest advertising. In the past, Researchers have worked on different models to detect clickbait, the majority of them used machine learning algorithms. This research proposes a method for detecting clickbait using Bi-LSTM. In this research, We have created a dataset using multiple sources like Twitter, ViralNova, Buzzfeed, TheOdyssey, Thatscoop, and various news sites. Moreover, We used Bidirectional Long Short Term Memory (Bi-LSTM) and compared it with other deep learning techniques like Gated Recurrent Unit (GRU), Dense Model, and Long Short Term Memory. Remarkably, Bi-LSTM outperformed all the deep learning models applied. Additionally, We evaluated how well our proposed approach performed against other currently used methods. In Machine learning classifiers, SVM, and in Deep learning models, Bi-LSTM has achieved almost 93% accuracy and Bi-LSTM has a better recall value.