Automated Sentiment Analysis of Natural Language Text using Machine Learning

Abstract

This paper presents an approach toclassify sentiment in peer reviews of papers submitted in Journals and Conference such as prepublication peer reviews, written before the paper is published and post publication peer reviews, written after the publication.Although the peer reviews are highly technical but they also contain sentiments. The proposed approach performs automatic sentiment analysis and polarity (Negative, Positive) classification of peer rev iews of scientific papers. Our approach finds the sentiment strength of word in the sentence by using term frequency and inverse document frequency weighting and then uses a Markov Logic based algorithm to assign weights to sentiment words based on their strength according to SentiWordNet3.0 and TFIDF weights and calculating the overall sentiment polarity of the sentence. The results of the used approach prove that our approach is comparatively more affective and accurate as compared to similar approaches.