Determining customer satisfaction dimensions in airlines via sentence embedding-based topic modeling

In the highly competitive aviation industry, ensuring customer satisfaction and converting passengers into loyal customers is considered a key competitive advantage. Understanding how passengers evaluate the services provided and identifying the factors that contribute to customer satisfaction are important for airlines. Alongside conventional approaches, newly developed analytical methods now offer opportunities to examine online reviews of passengers' travel experiences. In this context, the present study aimed to analyze the factors determining customer satisfaction in airline services based on online passenger reviews. A total of 436,343 reviews from various global airline carriers were analyzed, and contemporary text mining techniques were employed to identify key dimensions of customer satisfaction. Firstly, a performance test was conducted to compare traditional and sentence-based topic modeling approaches and to determine the most suitable method for the dataset. CombinedTM, a sentence embedding-based model utilizing pre-trained language representations, demonstrated superior performance across selected evaluation metrics and was adopted for the subsequent analysis. The findings revealed that customer satisfaction dimensions revolve around various service attributes such as delays, pricing, cabin class, seating, airport procedures, customer service, value for money, cabin crew, and in-flight services. Furthermore, the study offered a detailed analysis of satisfaction dimensions in relation to different airline business models (full-service vs. low-cost carriers), cabin classes (first, business, and economy), and travel types (domestic vs. international). In conclusion, the study contributes to the literature by demonstrating the applicability of sentence embedding-based topic modeling and offers practical insights for both researchers and industry practitioners.