A Computational Topic Modeling Approach to Airplane Boarding Literature: LDA vs. BERTopic Analysis

The airplane boarding process has retained cardinal attention within the civil engineering field due to its potential in reducing the costs generated for the airlines and improving customer satisfaction. The airplane boarding process has been a central point in mitigating the transmission of COVID-19, making it vital in the aircraft turnaround operation. Given their strong correlation with airport activities, over time, only some of the boarding methods proposed in the scientific literature have been implemented by airlines. The academic community simulated numerous boarding process methods, including simulation-based methods such as cellular automata and agent-based modeling, analytical and mathematical modeling, optimization and operations research approaches, or the growing field of data-driven and machine learning methods. The present paper aims to offer a common background knowledge on the works published in the field from the perspective of the approached topics. As a result, 76 papers have been extracted from Clarivate Analytics’ Web of Science platform and curated through the guidelines offered by the PRISMA approach, keeping only 73 for analysis. Latent Dirichlet Allocation (LDA) and BERTopic have been used for the extraction of topics by considering titles, abstracts, and keywords. Four major thematic streams in the airplane-boarding literature, across each type of analysis, with a high degree of overlapping in terms of results. It has been noticed that the largest body of work focuses on simulation-based analyses of boarding strategies, passenger interference, congestion dynamics, and boarding under operational and health-risk constraints, particularly in the context of social distancing and apron-bus or remote-stand operations. Additionally, another topic revolves around heterogeneous passenger speeds and queue dynamics, while another topic refers to experimental studies on perceived and subjective boarding time. A critical comparison of the extracted topics is provided with the purpose of better shaping the research conducted so far in this field.