Sentimental Analysis of E-Commerce Product Reviews for Product Specific Insights

The abundance of digital data available in today’s environment can improve quality of life and offer helpful insights. In the e-commerce space, customer evaluations and comments about a product may truly be utilized to gain valuable information for the seller, which can then be applied to the product’s enhancement. This study seeks to collect comments on the product’s surrounding features as well as the polarity of the sentiments expressed in product reviews. By doing this, the vendor will be able to comprehend the consumer market, consider ideas, and adjust their product properly. The model implementation consists of three parts, aspect term extraction, opinion term extraction, and sentiment score calculation. An unsupervised rule-based model has been implemented to extract aspect and opinion terms, whereas a BERT model to obtain the sentiment score.