This project presents an approach for sentiment analysis to mine sentiments of user reviews by classifying them as per their corresponding ratings along with a graphical visualization of the tests and results. We extract features from text reviews and use different models to predict review ratings. By using unigrams and bigrams as features, we build a rich predictor that identifies phrases and their polarities and how they could predict product ratings. The implementation makes use of Python and its libraries for Natural Language Processing and Data Visualization. The results of our work can be used to better understand customer feedback by ecommerce companies to mitigate the factors that lead to low ratings and to optimize the factors that customers appreciate. We put forward our approach along with its implementation, evaluation, visualization and conclusion.
