Aspect-Based Sentiment Analysis of Global, National, and Local Coffee Consumer Perceptions on Google Reviews Using a Python Model

This study investigates consumer sentimentstoward three tiers of coffee brands: global, national and local,using an aspect - based sentiment analysis (ABSA) of GoogleMaps Reviews. Data collection and analysis were conducted on adataset of 3,034 reviews in Bahasa Indonesia from one brand foreach tier. The research employed a nine - stage Python pipeline,starting from text preprocessing and followed by a rule - basedapproach for aspect extraction, followed by lexicon - basedsentiment analysis, feature creation, model building and trainingwith 5-fold stratified cross - validation. Five aspects wereanalyzed, as perceived by consumers: ambiance, packaging,price, service and taste. The research examined and comparedseven supervised machine learning classifiers, includingIndoBERT, SVM, Random Forest, Naive Bayes, LogisticRegression, Gradient Boosting and Decision Tree, underidentical conditions, evaluating each about its accuracy alongwith weighted and macro - averaged precision, recall and F1-score. IndoBERT achieved the highest outcome with 0.9275accuracy and 0.9275 weighted F1 and 0.8449 macro F1 metrics,closely trailed by SVM with 0.8902 accuracy and 0.8902 weightedF1. Cross - tier analysis revealed generally positive sentimentstoward all three categories of coffee brands. Global brands werefavored for their ambiance and taste but criticized for price andinconsistent service. National brands also had positive sentimentsfor ambiance and taste, yet there were observations ofinconsistencies among outlets and long wait times for customers.Local brands were positively associated with their ambiance andtaste but had some drawbacks, such as limited seating due tooperational constraints. These findings may contribute to thedevelopment of more effective e - business strategies, improvedcustomer relationship management and the designing of digitalcustomer experiences within the Indonesian coffee retaillandscape.