Improving sentiment analysis efficacy through feature synchronization

Social media platforms are becoming a rich source of valuable information through sharing and publishing user generated reviews and comments. The identification and extraction of subjective information from a piece of text is a crucial challenge in sentiment analysis. Numerous techniques have been proposed that aimed to analyze the sentiments of the text. However, accuracy was compromised due to inadequate identification of intensity, sentiments shifter and negation of words as well as divergence between tweets and their associated labels. In this study, Prescriptive Sentiment Analysis (PSA) based on features synchronization has been introduced for increasing accuracy in text sentiment analysis. At first, pre-processing has been performed which includes removal of stop words and tokenizing the text into sentiment words, intensity words, sentiment shifters and negation words. Secondly; polarity of intensity words, clauses and sentiment shifters in the text are calculated. Identification and removal of ambiguity between extracted features and their associated labels have been accomplished through feature synchronization. The K-Nearest Neighbor (KNN) has been implemented to predict text trend based on synchronized features. The proposed approach has been evaluated on publicly available datasets of twitter and movie reviews. Experimental results show significant improvement in sentiment analysis efficiency as compared to other baseline methods.