Sentiment analysis is a valuable knowledge resource to understand collective sentiments from the Web and helps make better informed decisions. Sentiments may be positive, negative or objective and the method of assigning sentiment weights to terms and sentences are important factors in determining the accuracy of the sentiment classification. We use standard methods such as Natural Language Processing, Support Vector Machines and SentiWordNet lexical resource. Our work aims at improving the sentiment classification by modifying the sentiment values returned by SentiWordNet for intensifiers based on the context to the semantic of the words related to the intensifier. We also reassign some of the objective words to either positive or negative sentiment. We test our sentiment classification method with product reviews of digital cameras gathered from Amazon and ebay and shows that our method improves the prediction accuracy.
J. Bhaskar, Sruthi, K., and Prof. Nedungadi, P., “Enhanced sentiment analysis of informal textual communication in social media by considering objective words and intensifiers”, in IEEE International Conference on Recent Advances and Innovations in Engineering (ICRAIE), 2014, Jaipur, 2014.