Incorporating semantic knowledge for sentiment analysis

Abstract

We report work on using knowledge of senti- ment-bearing words in statistical approaches to automatic sentiment analysis and opinion mining (SA & OM). Our main contribution lies in constructing document feature vectors that are sentiment-sensitive and use word knowledge. This is achieved by incorporating sentiment-bearing words as features in docu- ment vectors, extracted with the help of Sen- tiWordNet which is essentially the wordnet with sentiment scores attached to the synsets. Support Vector Machines (SVM) have been used to classify documents into positive and negative polarity (i.e., sentiment) classes. Ex- periments show that we achieve state of art performance in sentiment analysis of standard movie reviews dataset and locally created product review dataset.