Statistical feature based approach for aspect oriented sentiment analysis

User generated data is tremendously increasing on the web. Many social networking sites are popular and powerful platforms for finding friends and people like to post comments. This data can be exploited for analyzing sentiments of the user. Feature extraction from text is crucial stage for sentiment analysis. This paper is intended to propose a simple method for feature extraction from text. Proposed method measures semantic relatedness /similarity from WordNet. Advantage of the proposed method is that it is domain independent. SentiWordNet 3.0 has been used for polarity classification. Average accuracy obtained for camera domain is 0.847 and restaurant domain is 90%.