Analysis of online reviews of airline passengers through text mining

Passenger reviews on social media mostly consist of text data. These text data contain valuable but latent information, which helps to understand passengers better. Using text mining techniques, this study reveals the aspect based sentiments of airline passengers toward airline service attributes in online reviews shared through passengers' flight experiences, and explains the relationships of these sentiments with value for money, satisfaction and recommendation behavior. To this end, 42,881 passenger reviews of 50 full service network carriers on the Skytrax platform were collected. Using the Latent Dirichlet Allocation method, the service attributes mentioned in the reviews were determined as comfort, cabin staff, flight experience, ground handling, catering and in-flight entertainment. Aspect-Based Sentiment Analysis method was used to determine passenger sentiments toward airline service attributes. Logistic regression and linear regression analyzes were also performed to examine the relationships between passengers' value for money perception, satisfaction, recommendation behavior and passenger sentiments by using service attribute based passenger sentiments data. In all regression models, it was revealed that positive passenger sentiments affect the dependent variables positively. In the context of passenger sentiments, cabin staff was defined as the most important service attribute for all dependent variables. Passenger sentiments toward comfort and ground handling were other prominent service attributes in the regression models. As a result, it has been revealed that passenger sentiments toward airline service attributes have significant effects on value for money, satisfaction and recommendation behavior.