Sentiment Analysis On Amazon Product Reviews

With the explosion of social networking sites, blogs and review sites a lot of information is available on the web. This information contains emotions and opinions about various product features and the makers of these products. This form of opinion and feedback is important to the companies developing these products as well as the companies that want to develop better rival products. Sentiment Analysis is the task of analyzing all this data, retrieving opinions about these products and services and classifying them as positive or negative, in other words good or bad. The key parts of any review of any product are the numeric rating and the textual description provided along with this product. In our project we will take into consideration both these vectors for product reviews to conclusively decide on a classifier that is best suited to analysis of product reviews. We have gathered reviews and based on the features that best describe the sentiment for each review, we have created a feature set of 1000 features, and with this limited set we will determine which classifier gives the best result on review type data. To determine the best classifier we perform evaluations on it, by running various data set generators, calculating the resubstitution and generalization errors for each classifier. We then use the mean of these results to compute the paired Student’s t-test to relatively compare the performance of the classifiers. Based on the results of this evaluation, we can state which is the best classifier.