The e-commerce sector is increasingly developing consumers innovative services such as product reviews used to meet consumer needs. Product reviews allow online shoppers to share their opinions regarding a product. Online reviews have an immense impact on a customer's decision to purchase because these opinions are based on another customer's experience with a specific product. Consequently, an e-commerce site evaluates such reviews to find consumers' sentiments about the product. Therefore, this project proposes a feature-specific approach to online reviews using topic modeling and sentiment analysis. 

The proposed approach is to foster feature-specific sentiment analysis that provides a convenient product review for customers instead of the conventional and unstructured review texts. Sentiment analysis involves drawing meaning from text using computer programs to extract consumer sentiments about an online product (Ali et al., 2019). Topic modeling is a statistical application used to identify abstract topics in a collection of texts or documents to learn important patterns of words (Ali et al., 2017). When an online shopper is reading product reviews, some can be tedious to comprehend in a short time. The feature-specific approach uses semantic analysis and topic modeling to analyze numerous reviews and give precise topic texts summarizing lengthy reviews. 

In this research, topic modeling and sentiment analysis are implemented in Python using the open-source APIs: Gensim and Textblob, respectively. The project uses an advanced natural language processing algorithm to create a feature-specific sentiment analysis for product reviews. The topic words that are extracted will be mapped, thus allowing feature-specific sentiment analysis on the product reviews. The study uses Amazon's Echo Dot dataset obtained from Kaggle because Amazon's platform has sufficient product reviews and related metadata that span a period. The dataset entails product metadata such as brand, price, and category information and reviews such as text, helpfulness votes, and ratings.