Identification of Clickbait Using Gated Recurrent Units and Improving Efficiency through User Ratings

Online news and newspaper headlines use catchy phrases or statements to tempt the readers to read a particular article. These are called clickbaits, which can often be misleading. The information provided is just enough to arouse the curiosity among the readers which later leads to bad user experience. Apart from being misleading these can instigate many activities unfit for the society. They lead to user dissatisfaction and increase in mistrust in the journalism process. This is the need of the hour to detect these clickbaits precisely and there are several machine learning methods being used to do so. Our approach uses a deep learning method using Gated Recurrent Unit (GRU) to identify the clickbaits in the given dataset. Built upon that is an application that can be used to distinguish between a clickbait and a non-clickbait headline and a list of news articles that the user can rate as in a range from clickbait to useful which will help the model learn better. Our improvised model gave us an accuracy of 89%. Clickbait is different from spam and fake content on the web, more research work is being done to identify low quality content.