An Approach to Product Rating based on Aspect Ranking in Opinion Mining

Objectives: The main objective of this paper is product ranking framework that automatically generate from the product review given by own individual user’s opinion by sentiment analysis. The additional work conducts product ranking that is spawned in a source of a URL in a large scale. This aspect ranking from the products is used for analyzing the feedback in online marketing websites and rating of a product. Methods/Analysis: Bag-of-words (BOW) is the popular way to analyze the text and perform sentimental analysis. Our proposed system focuses on unigram, bigram and neutral polarity. Our proposed system classifies the polarity of each opinion words. In this the dependent and independent word will be given as input and stored in the database. We classify the meaning of the word, in some cases these words will give different meaning or contrast polarity. So it is very important to train the polarity of the words accurately. The classifier is used to predict the sentiment on each aspect by analyzing the positive and negative words separately. Finally aspect ranking is performed to obtain the final graphical result. Most of the time, the sentiments are classified as positive or negative. There are also some neutral words which should also be categorized. So, it’s important to train the classifier to predict the positive, negative and neutral words. Finally, aspect ranking of the features is performed to rate the products. Findings: This paper has analyzed and categorized the reviews as positive, negative and neutral. It is found that the neutral reviews which are omitted in most of the websites play an important role in deciding the rating of the product. Ten different products belonging to five different categories are considered from a popular website and the output is presented in the form of a 3D graph. Improvement: Numerous product reviews are available online in the internet. These reviews enable the users to choose between different products. Since, the reviews are many in number, valuable time of the consumer is wasted in going through each of the reviews and finally arriving at a conclusion. This paper aims at providing a snapshot of the overall product rating based on the key aspects which determine the product features. The ranking algorithm is used to rate the products in a scale of ten as an improvement to the star rating provided by most websites.