Developing techniques for finding the correlation between user reviews and userratings

Abstract

In order to predict how a user will respond to a product, we mustuncover the tastes of the user and the properties of the product. Forexample, in order to predict whether a user will enjoy a food or not it depends on theuser‟s level of interest in which type of foods user likes most. User feedback is required todiscover these dimensions, which comes in the form of ratings andreviews. In this thesis, we aim to find a correlation between user reviews and star ratings. However, traditional methods discard review text, whichmakes these latent factors difficult to interpret.In this thesis, Firstly, our approach is to analyze a text review according to the texts and find the sentiment of the user review whether it is a positive review or negative review. Secondly, after getting the sentimental analysis,finding the correlation between the sentimental analysis and user star ratings. our approachmore accurately predicts sentiments by analysing the informationpresent in review text. Our discovered methodology can be used to identify the useful and representative reviews.