Bengali Document Categorization Using Deep Learning Approach

Abstract

The rapid growth of social media and microblogging sites not only gives places for increasing free expression and individual voices, but also allows people to engage in antisocial conduct such as online harassment, cyberbullying, and hate speech. Several initiatives, mainly for highly resourced languages like English, have been proposed to leverage this data for social and antisocial behavior analysis, document categorization, and sentiment analysis by predicting scenarios.But when it comes to sub-sided languages such as the Bengali, Hindi, Urdu and many others, the researchers in the outgrowing field of Natural Language Processing suffers from a great amount of deal because of the lack of basic components and materials. In the case of our experiments, we have used a dataset of news data consisting of a total of 19137. The CNN-BiLSTM deep learning approach was used in the case of categorizing different classes. The main purpose of this work was to determine between the classes which could be helped in order to help the user’s concussion to help individuals to identify in which categories the data resembles.