Sentiment analysis of islamic news data using hyper-concepts

Abstract

Sentiment Analysis is the extraction of writers feeling from a written manuscript. It aims at predicting the sentiment of a particular text using automated means. There are two main ways to make predictions: (1) lexicon-based techniques; (2) machine learning based techniques. Our paper contributes in the machine learning techniques, which are more accurate. Furthermore, we have adopted a hyper-conceptual method as our primary feature extraction technique. This method extracts the keywords in a hierarchical ordering of importance. Classification is then performed using the Random Forest classifier that predicts the sentiment of each document. We were able to obtain an accuracy of 90% on an comments collected from Al Jazeera News website.