Investigating Machine Learning Techniques for User Sentiment Analysis

Related Work

In movie review domain where the application of sentiment analysis has the greatest potential to generate value, it is an interesting and challenging testing ground for different sentiment classification approaches. It was hypothesized that this was due to the tendency of reviewers to rate the individual elements of items differently from the item as a whole within the same review. (Bo Pang et al., 2002) suggest, the machine learning methods and features used when classifying movie reviews do not have to be specific to that domain but its benefits may easily transfer to other areas where sentiment classification can be applied. For binary (positive and negative) sentiment classification, (Turney et al., 2002) proposed the positive and negative terms can be counted and expressions in a review used to determine its polarity. (Kennedy et al., 2006) who had considered negation words, intensifiers, and diminishes taken into account and managed for improving the accuracy of the system. In this article Support Vector Machines (SVMs) method was proposed, and it was inferred that this machine learning algorithm performs significantly better than the term-counting method by considering negation words. (Pang et al., 2002) and (Pang et al., 2004) have compared the performance of various classification algorithms when determining the sentiment of a document, and also proved that SVMs were generally the best approach. Unigrams, bigrams, part-of-speech (POS) tags and term positions were considered as features, and using unigrams gives the best results. When using very simple features with multi-aspect sentiment analysis the SVM classification algorithm can be effective for performing sentiment analysis.