A Machine Learning Approach to Aspect-Based Sentiment Analysis of Hotel Reviews

With the increasing reliance on online hotel reviews for decision-making, understanding customer feedback at a granular level has become essential for hospitality stakeholders. This research presents a machine learning approach to Aspect-Based Sentiment Analysis (ABSA) of hotel reviews using the TripAdvisor dataset. The methodology integrates advanced preprocessing techniques such as lemmatization, stopword removal, and contextual tokenization, followed by training multiple classical machine learning models (Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, XGBoost, and SVM) and deep learning models (LSTM, BERT). Furthermore, we explore aspect-level sentiment detection for key hotel service categories such as cleanliness, staff, food, and amenities using rule-based extraction and polarity scoring with VADER and TextBlob, alongside an attention-based neural model. Results demonstrate that deep learning approaches achieve superior performance compared to classical ML models. Finally, an interactive dashboard was developed to visualize sentiment trends, highlight frequently mentioned issues, and provide actionable insights for hotel managers.