Emoticon aware aspect based sentiment analysis of online product review

Abstract

With the rapid growth in e-commerce, reviews of popular products on the web have grown rapidly. When an individual wants to make a decision about buying a product or using a service, or when an organization wants to benefit by obtaining the public opinion or to market its products, identify new opportunities, predict sales trends, or manage its reputation, they have access to a huge number of
user reviews but reading and analyzing all of them is a tedious task. If someone reads only few numbers of reviews and comes to a decision, then the decision could be biased. Because of these reasons having a better data mining technique to mine these product reviews, which are in semi structured format is very important. Therefore, there is a growing need to analyze and summarize a
large collection of reviews automatically to overcome subjective biases and mental limitations. With sentiment analysis techniques, it is possible to analyze a large amount of available data, and extract opinions from them that may help both customers and organizations to make decisions. However, incorporation of emoticons in sentiment analysis of online product review has received little attention,
and there has also been the problem of misspelled words, implicitly mentioned aspects and emoticon having a different idea with its associated text in a review. In this research work, emoticon aware aspect based sentiment analysis model is proposed, as incorporation of emoticon in sentiment analysis is likely to give complete and accurate results. The proposed model was evaluated using dataset of
2085 iPhone mobile reviews downloaded from Amazon website. Emoticon Lexicon Technique, POS Tagger and SentiWordNet were used for the successful implementation of the Model using RStudio Software Packages. The result of this study shows that the usage of emoticons, and consideration of implicitly mentioned aspects improves the performance of the model with overall accuracy of 88.5%,
Precision of 88.1%, and recall value of 84.6% compared with results of previous work.