Theme extraction based on NLP

Abstract

Theme Extraction is a method of identifying, evaluating, and understanding human perception about a product in form of key features or themes that we extract dynamically from a given set of reviews within a data set. These themes are then categories to form an opinion about a given feature inside of the product through which we can analyze the advantages as well as the short comings of a given product or organization. These key features are then displayed on to the user for them to make a wise decision based on their likes and dislikes which they can compare with the user base that have already formed a review. All of this happens seamlessly with the help of Natural language Techniques that enables us to dynamically extract features or themes and generalizes an opinion score alongside it to represent thousands of reviews in a small concise manner. To make this happen, we consider seven different steps: (i)Text Pre-Processing, (ii) Removal of Stop-Words, (iii) Vader Sentiment Analysis, (iv) Feature Extraction using HAC, (v) Classification of Key Features using MOS, (vi) Testing the Accuracy of the Score and (vii) Creation of Word-Cloud using Features.