Sentiment analysis for Bengali newspaper headlines

Abstract

In this project, we propose a system that assigns scores indicating positive or negative to newspaper headlines. Many works have been done on sentiment analysis, document clustering for newspaper headlines in English. We are going to do the same for Bengali language. News headlines from one Bengali newspaper [1] is used for the purpose of the project. As there is no dataset of headlines for this newspaper, we are using web crawler to get necessary headlines to make a dataset to use for this project. For our proposed system, a number of classifiers will be used such as Support Vector Machine [2], Logistic Regression [17] etc. Through the experiments, our aim is to establish a system, which can identify positive and negative news accurately.