Eco‐Conscious Preferences, Emotional Attributes, and Polarity in Sentiment Analysis of Organic Food App Reviews: A Feature Engineering and Deep Learning Approach

This study investigates the impact of eco‐conscious preferences, emotional attributes, review score, and review polarity on overall sentiment in user reviews for the Thrive Market app, an e‐commerce platform for organic and natural food products. Grounded in the theory of planned behavior (TPB), emotion‐centric theories, and expectation‐confirmation theory (ECT), the study identifies key predictors of sentiment. Data from 7012 valid reviews were preprocessed and analyzed using feature engineering techniques to extract eco‐conscious preferences and emotional factors based on the NRC emotional lexicon. Sentiment analysis was conducted using the SentimentIntensityAnalyzer, and a multilayer perceptron (MLP) deep learning model was employed to predict sentiment. The MLP model achieved an accuracy of 92%, with particularly high performance in predicting positive sentiment. The results indicate that eco‐conscious preferences, review score, polarity, and emotional attributes like joy, trust, and anticipation have a positive impact on sentiment. In contrast, emotional attributes such as sadness, anger, fear, and disgust are negatively associated with sentiment. The findings highlight the significant role of both emotional factors and eco‐conscious preferences in shaping consumer sentiment, offering actionable insights for marketers in the organic food sector.