This study evaluates sentiment analysis techniques for assessing tourist experiences with Airbnb accommodations in Albania by comparing textual reviews with conventional numerical ratings. Since star ratings may not fully capture the contextual and subjective aspects of customer experiences, Natural Language Processing (NLP) techniques are applied to extract sentiment from textual feedback. The study compares a BERT-based Transformer model with traditional machine-learning methods, including Support Vector Machine (SVM) and Naive Bayes. The experimental results indicate that BERT provides higher sentiment-classification performance, demonstrating its ability to capture contextual and semantic information in tourist reviews. The findings highlight the potential of BERT-based sentiment analysis to complement numerical ratings and provide more detailed insights into tourist satisfaction and accommodation service quality.

Sentiment analysis BERT NLP transformer data preprocessing removing noise truecasing stopwords lemmatization text normalization