Sentiment Analysis and Similarity Evaluation for Heterogeneous-Domain Product Reviews
Mangal Singh, Md Tabrez Nafis, Neel Mani · International Journal of Computer Applications · 2016
Sentiment analysis and classification is a prominent research topic in academics as well as in industrial field.Since each customer reviews text always state emotion about a target domain, sentiment classification is a highly domain dependent task and present study considered the reviews from heterogeneous domains.Generally researchers classify the customer review with positive, negative and neutral sentiments but a positive review can be highly positive and a negative review can be highly negative, so sentiment analysis about a review can be more effective if a sentiment scale is also defined for such greater degree of positivity or negativity.We defined a framework to classify heterogeneous product reviews with degree of polarity on a sentiment scale of range -2 to 2. For each review, an intermediate form is calculated using sentiment vectors which is further processed to calculate the sentiment polarity magnitude and similarity of reviews.