The subject of this repository was to demonstrate a very basic sentiment analysis for product reviews. A multiclass classification task was also performed to identify emotions behind on customer reviews. Sentiment analysis shows to be important and beneficial for business usage. Regularly performed on text-based information can assist organizations with checking item opinion in customer feedback and comprehend customer needs. E-commerce platforms, such as Amazon, gives its customers the option of providing ratings and comments as feedback for its products. While numeric ratings are usefull, it has been observed that the actual sentiment of the customers is better reflected in their comments. Sentiment analysis is a type of text mining useful to uncover customer opinions and understand the general customer sentiment about a product. This field is continually advancing and creating.
The sentiment scoring was completed using two models: VADER and RoBERTa (transfer learning). A classification was also performed using the stochastic gradient descent (SGD) model. The evaluation returned overall 0.9110 accuracy. We proved that sentiment classification is possible on this dataset of instrument reviews.
