Sentiment Analysis on Amanzon Reviews using Machine Learning

Abstract

With the advent of social media, people are now more comfortable than ever to express their thoughts, opinions, and emotions online. The proliferation of these comments, whether positive or negative, makes it crucial to analyse them accurately in order to grasp the true intentions of the writer. To achieve this, sentiment analysis is used to decipher the perspective of the text. In our study, we propose a novel approach that takes into account the sentimental aspects of the item being reviewed. To validate our approach, we utilized Amazon consumer reviews, specifically the Amazon musical Instruments Reviews dataset collected from the Kaggle repository by Eswar Chand. In this dataset, user ratings were initially detected in each analysis, after which we conducted pre-processing operations, such as creating a sentiment column, tokenization, reviewing text-punctuation cleaning, and eliminating stop-words to extract meaningful information such as the positivity or negativity of the feedback. Our main goal was to analyse this data on an aspect level, which would be highly beneficial to marketers in comprehending consumer preferences and adapting their strategies accordingly. Furthermore, we also provide insights into possible future work for text classification. Ultimately, our study presents a new approach to sentiment analysis that can enhance our understanding of online feedback and facilitate more effective marketing practices.