Filtering Bengali Political and Sports News of Social Media from Textual Information
Lutfun Nahar, Zinnia Sultana, Nilufar Jahan, Ummay Jannat · 2019 1st International Conference on Advances in Science, Engineering and Robotics Technology (ICASERT) · 2019
Web is being updated rapidly day by day. As a result of the modern media development, news also published very quickly. Generally, it takes less time to appear in the newspapers than an incident happens. But searching the relevant news through internet about one's choice is the foremost task. News classification is the promising area that fulfills what they want. As Bengali is the common language for a huge number of people; they feel comfortable to use it for publishing news in different news portal and social media. Different post and comments are also written in Bangla. So Bangla new in web are promptly increased, classification of Bangla news according to the subject is definitely needed. This paper focused on some machine learning methods such as Naïve Bayes Classifier that is based on probability theory, Support Vector Machine and Neural Networks to classify news as well as post or comments written in Bangla. 1000 long written post are collected from different news portal as well as social media. A Self- created News Corpus is used for experimental analysis. For training and testing a Feature Vector is constructed which is based on TF-IDF. The result shows that the Naive Bayes performs better than other three algorithms. In order to get better accuracy stop word are removed from documents. Bengali stemming has done to reduce the size of the dictionary. Moreover a comparison has been made to show the accuracy of different algorithm.