Abstract
This study addresses the challenge of analyzing complex visitor experiences through a novel Generative Pretrained Transformer (GPT)-based Sentiment Analysis Framework that bridges theoretical and methodological gaps in artificial intelligence (AI) applications for experience evaluation. The framework integrates the complementary Experience Economy and Service Quality (SERVQUAL) theories, developing a approach to understanding visitor experiences through advanced natural language processing and machine learning techniques.
Using the British Museum as a case study, we analyze 1814 Google Reviews from January to July 2024, achieving high validation metrics (accuracy: 0.89, F1 score: 0.87, Cohen's Kappa: 0.76). Our framework synthesizes five core dimensions: Museum Exhibitions and Artworks, Architectural Design and Atmosphere, Emotional, Intellectual, and Social Experience, Staff Services, and Amenities through systematic prompt development and aspect-based sentiment analysis (ABSA) implementation.
Results showed distinct satisfaction patterns across dimensions (exhibitions scoring 0.82, staff services 0.20) and reveal a significant seasonal decline in overall satisfaction from 4.70 in January to 4.42 in July, with AI-enabled insights identifying specific service delivery challenges. This research makes three key contributions: (1) establishing a theoretical foundation that integrates service quality and experiential elements for comprehensive visitor experience analysis through AI applications, (2) developing a reproducible methodology that bridges quantitative sentiment scores with qualitative insights using advanced machine learning, and (3) providing actionable recommendations for practical implementation.
Beyond museums, this framework can be adapted to other experience-centric sectors including hospitality, tourism, performing arts, and theme parks, offering a scalable solution for organizations requiring sophisticated analysis of customer feedback across diverse experiential contexts.