Performance Measurement of Multiple Supervised Learning Algorithms for Bengali News Headline Sentiment Classification

The reading newspaper is a common habit in today's life. Before reading news article all are focused on the news headline. Understanding the meaning of news headline everybody can easily identify the news types. That means the containing news article provides positive or negative news. Analysis of the sentiment of the news headline is a good solution for this kind of problem. Sentiment Analysis is a chief part of Natural Language Processing. It mines any kinds of opinion and set the sentiment of any text. We proposed a method for Bengali news headline sentiment measurement with different kinds of the supervised learning algorithm and their performance. Firstly, we set sentiment of each news headline then used the classification method to predicting the news headline which was containing a positive or negative headline. After all, Bengali is one of the most used languages in this world. A lot of research work done previously in a different language but very few in the Bengali language. So, increasing the Bengali language research resource need to develop different kinds of tools and technology.

I. Introduction

Any human language problems are solved by NLP in AI research fields. It grasps the concept of human language problems and tries to provide a solution for the machine. The machine learning algorithm is the most usable algorithm for understanding the NLP problem with the solution. Machine learning is a concept which meaning an automatic learning system. A few approaches have in machine learning such as supervised, unsupervised and semi-supervised learning. In supervised learning provided labelled data with input and output but in unsupervised learning provided only unlabeled input data and out will generate from input data. Semi-supervised learning is made from both combinations where the label and unlabeled data have in mixed.