Sentiment and Complaint Analysis of Airline Tweets via Natural Language Processing

Abstract

Airline customers often share their experiences on social media, from glowing reviews to complaints about delays, cancellations, or lost luggage. In this study, we analyze a collection of airline-related tweets labeled by sentiment and, for negative tweets, by the specific reason for dissatisfaction. We explore the data with word clouds, train a DistilBERT model to classify sentiment and negative-reason categories, and use a multinomial logit model to highlight words most associated with positive or negative feedback.