Mohamed Assoud

This paper proposes and illustrates an explainable Natural Language Processing (NLP) pipeline for detecting and classifying brand hate in online consumer reviews. Moving beyond traditional sentiment analysis approaches that reduce consumer feedback to simple positive–negative polarity, the study adopts a hybrid lexicon-based framework that combines sentiment scoring, emotion profiling using the NRC Emotion Lexicon, and rule-based classification to operationalize the multidimensional nature of brand hate. The pipeline is implemented as an interactive R Shiny application and follows a structured workflow encompassing automated review scraping from ConsumerAffairs, sentiment-based filtering of negative content, textual preprocessing, emotion analysis, and a two-stage classification procedure distinguishing Mild, Moderate, Strong, and Hybrid forms of brand hate. To enhance transparency and interpretability, analytical outputs are communicated through interactive visualizations, including hate intensity distributions, reason-based breakdowns, emotion word clouds, and sentiment–intensity bubble plots. The artifact is demonstrated through a step-by-step application using real consumer reviews, serving to document system behavior and analytical logic rather than to test theory or generalize empirical findings. By translating established brand hate theories into a computationally accessible and explainable tool, this study contributes a methodological foundation for future large-scale, longitudinal, and cross-platform investigations of consumer hostility, while offering managers a practical means to monitor and understand emerging brand hate dynamics.