Sentiment Classification using Cross-Domain Data with N-Gram Method
Sentiment classification is a data mining method that widely different from other types of traditional information extraction. It includes only two basic categories (positive/negative and stars) compare to text classification. Sentiment analysis is one of the most prominent case of natural language processing. Sentiment classification will predict the polarity of reviews automatically either positive or negative with respect to sentiment polarity of sentence. This paper presents a classification method extended with feature clusters using n-gram words as features. Using a Spectral Feature Alignment algorithm those n-gram features are aligned as domain specific and domain independent for feature clustering. Domain independence achieved using cross domains. N-gram features can provide much better results for polarity classification within smaller false positive and false negative rates.
Keywords - Cross domain, sentiment analysis, Feature Extraction, Classification, Machine Learning