A review of feature selection in sentiment analysis using information gain and domain specific ontology

There is a continued interest in understanding people’s interest through the contents they share online. However, the data generated is massive, characterized by textual jargons and tokens that contain no sentiment or opinion value. One way of reducing the data dimension and pruning of irrelevant features is feature selection. However, the existing approaches of feature selection are still inefficient. Two prominent feature selection methods in sentiment analysis are information gain and ontology-based methods. Information gain has the disadvantage of not considering redundancy between features while ontology-based approach requires a lot of human intervention. The aim of this paper is to review these two methods. The review of these two methods shows that using the two methods in a two-step approach can overcome their limitations and provide an optimal feature set for sentiment analysis.