Aspect-Based Sentiment Analysis of Fashion and Apparel Related Reviews using Transformer Model

The increasing reliance on e-commerce has led to a surge in customer-generated content, making sentiment analysis a crucial tool for understanding consumer preferences. This study presents a novel aspect-based sentiment analysis (ABSA) approach to extract fine-grained insights from customer reviews across multiple product categories, including lingerie, women’s clothing, men’s clothing, kids’ apparel, home & furniture, and flowers & plants. We employ exploratory data analysis (EDA) to uncover key patterns in customer sentiment and utilize a transformer-based model with contrastive learning to enhance sentiment classification accuracy. Our approach effectively captures sentiment variations across nine different aspects, including quality, value for money, style, fit, material, warmth, comfort, support, and how well the product fits. The results demonstrate that our model outperforms traditional sentiment analysis methods, offering a more structured understanding of customer feedback. By analyzing customer sentiments from reviews, businesses can identify dissatisfaction patterns that may contribute to customer churn. Addressing these issues proactively can help improve customer retention and loyalty. These insights can help businesses refine their product offerings and improve customer satisfaction across diverse e-commerce categories.