Machine Learning Empowered Sentiment Analysis: Techniques and Insights

Sentiment analysis is a crucial area of contemporary research, particularly useful for examining text data and identifying sentiment elements. E-commerce platforms generate vast amounts of text each day through customer comments, reviews, tweets, and feedback. The rise of social networking sites has significantly enhanced communication and knowledge-sharing. Conducting aspect-based sentiment evaluations can provide businesses with valuable insights into consumer expectations, enabling them to adjust their strategies accordingly. Conveying the precise sentiment of a review can be challenging. This study introduces an approach focusing on the sentimental aspects of product characteristics. We analysed consumer reviews from Amazon and IMDB, using a dataset sourced from the UCI repository, which includes opinion ratings for each review. To derive meaningful information from the datasets and reduce noise, we performed pre-processing steps such as tokenization, punctuation removal, and elimination of whitespace, special characters, and stop words. For effective representation of the pre-processed data, we applied feature selection methods like term frequency-inverse document frequency (TF-IDF). We merged customer reviews from three datasets—Amazon, Yelp, and IMDB—before applying classification using algorithms including Naïve Bayes, Random Forest, K-Nearest Neighbour (KNN), and Support Vector Machine (SVM). Finally, we offer insights into potential future work in text classification.