Machine Learning Approach based Sentiment Analysis, Classification: An Application of Natural Language Processing

Abstract

Many social websites and android applications is flooded with user reviews and data in this modernized world. Every enthusiastic social person tends to express his or her views in form of comments and this is referred as sentiments. These comments not only express the peoples’ view but also able to know in depth knowledge mostly in case of any social media. Generally, these comments/reviews present in text format are unstructured in nature and hence text preprocessing technique is applied in prior to analysis. Natural language Processing is used to get the sentiments (positive, negative, neutral) with many feature extraction techniques. This paper analyses the Amazon product reviews with most popular feature extraction techniques and classification using machine learning approach for word and n-gram sentiment analysis levels. Term Frequency-Inverse Document Frequency (TF-IDF) with pruning is used for feature extraction and sentiment analysis then data mining decision tree based Random forest algorithm with feature weights is applied for sentiment classification. The work is implemented Rapidminer tool and suitable evaluation measure is used for assessing the performance of existing and proposed.