This report presents a comparative sentiment analysis of customer reviews for Rubbermaid and Windex products using two distinct approaches: a Generative AI (LLM-based) model and a traditional machine learning pipeline implemented in RapidMiner. The study evaluates how effectively each method classifies customer sentiment while examining their respective strengths in handling contextual nuance, sarcasm, and mixed-sentiment language.
The findings indicate that both approaches produce broadly consistent brand-level sentiment outcomes; however, notable differences emerge at the individual review level. Generative AI demonstrates superior contextual understanding and performs more accurately on nuanced or sarcastic reviews, whereas RapidMiner provides greater transparency through token-level explainability and deterministic scoring.
Overall, the analysis highlights the trade-off between contextual accuracy and model interpretability in sentiment analysis systems. The report concludes that Generative AI is better suited for complex consumer language, while RapidMiner-style analytical pipelines remain preferable for high-stakes, auditable, and compliance-sensitive applications requiring transparent and reproducible decision-making.