Multilevel Categorization of Bengali News Headlines using Bidirectional Gated Recurrent Unit

Text Classification is one of the popular tasks of Natural Language Processing. Like many other languages, Bengali is also gaining popularity in this field and the most famous work here is the categorization of different unlabelled news articles into some predefined categories like National, International, IT, and so on. This study focuses on classifying Bengali News Headlines instead of news articles from different online newspapers. It categorizes News Headlines into six different classes using Bidirectional Gated Recurrent Unit. The proposed model got an accuracy of 97% on training data and 84% on validation data which is quite satisfactory compare with the performances of other existing models in this field.