Sentiment Analysis and Fake Review Analysis of Amazon Products using NLP with R
The Objective of this study is to learn how different statistical NLP techniques can be used to analyze the Consumer Behaviour(Sentiment Analysis) and to check whether the consumer is real(Authenticity of Consumer Review Analysis).
My objective is not to classify each review as either positive or negative as I am more focussed on learning the different sentiments involved in the reviews for a product. This is why I have not used Supervised Machine Learning Algorithms.
There are two parts of this Study - 1. Sentiment Analysis 2. Fake Review Analysis
Sentiment Analysis focusses on determining whether the product has positive or negative or mixed reviews in general. It does not focus on defining each review as either negative or positive as I believe that the Sentiment Analysis is more beneficial if we try to analyze the consumer sentiments associated with a product and not focus on ananlyzing whether a product review was posotive or negative.
Fake Review Analysis focusses more on analysizing consumer behaviour when it comes to writing reviews. From a business stand point, identifying fake users is more beneficial rather than identifying fake reviews because we can remove or block the users so that no more fake reviews get added to any products.
- We were right in anticipating that the Amazon Products in general are very good.
- Both the most bought product - Amazon Echo Alexa 7 inch and the least bought product - Fire TV 4K have positive sentiments associated as per the text reviews.
The Objective of this Project is not to classify the reviews as fake or genuine but to identify the behavioural patterns of users which might be posting fake reviews. It is very difficult to classify one review independently as either fake or not. It is more beneficial to track one username because if it is identified as a bot then it is useful not just for one product but can be helpful for other products as well.