LLM-Assisted Semantic Mining of Passenger Perception in Humanistic Airport Waiting Areas: A Framework for Cultural Heritage Dissemination

Abstract

Airport waiting areas increasingly function as transport, cultural-communication, and informal public-learning spaces, yet spontaneous passenger perceptions are difficult to capture through preset questionnaires. This study develops a large language model (LLM)-assisted semantic mining framework that translates multi-source passenger text into traceable evidence for humanistic airport design. Reviews, social-media posts, service messages, and field interviews are cleaned and coded into five fields: spatial area, perceptual object, perceptual word, sentiment polarity, and design implication. The extracted semantics are organized into five dimensions: spatial scale, physical environment, humanistic atmosphere, functional facilities, and psychological emotion. Perception intensity, negative concentration, and a comprehensive shortcoming index support zone-specific diagnosis. Illustrative expressions such as "oppressive," "cold," "weak regional identity," "insufficient seating," and "waiting anxiety" are linked to circulation, environmental comfort, cultural integration, facilities, and emotional support. By connecting passenger language with spatial diagnosis and culturally responsive design strategies, the framework provides a concise methodological route for improving waiting-space experience and supporting cultural heritage dissemination.