Online Product Rating Estimation Using Aspect Level Sentiment Analysis

Abstract

newlineABSTRACT newlineInternet provides opinionated content which can influence users. Reviews on product can highly affect the user s opinion while purchasing that particular product or related one through online shopping sites. However, the information that is provided by the seller is not likely to trust upon as much the information provided by buyers on any product. The information that a customer provides on a product can help other buyers to make an efficient decision about the product. Considering the developer and producer s point of view on the product, it can be helpful for them to know the reviews or opinions of customers on their products because it can help them in knowing deformities in product and improvement areas for the same. This information can help them efficiently develop the products and managing the same proficiently. The work presented in this thesis is based on bi-polar word adjustment for optimizing aspect level sentiment analysis on online reviews. For the learning approach, the ratings on the reviews have been utilized. Online reviews are inherently biased means people provide rating which can be different from what they are writing which can produce the wrong sentiment rating of a given aspect of a product easily. For an accurate aspect level sentiment analysis, the bi-polar words based on context which can affect the positive or negative polarity of the aspect has been taken into consideration and optimized the results. The model implementing bi-polar Aspect Level Sentiment Analysis utilizing multiple kernels of SVM i.e. Support Vector Machines have been trained. First, Lexicon Aspect Level Sentiment Analysis has been done and then the Bipolar Aspect Level Sentiment analysis has been performed with bipolar word adjustment for optimization and efficient results. newlineThe work includes a collection of data in which sentiments as data are collected, then the analysis of the collected data is to be performed and finally, evaluation on it is to be made.