A deep-neural associative touristic experience recommendation approach with real-time context injection

Abstract Touristic experience (TE) is a unique and underexplored category of digital tourism products. It is multi-modal, dynamic, and highly subjective, which poses significant challenges for traditional tourism recommender systems. To address these challenges, we present an Intelligent Embedded Recommender Network (IERN), a framework specifically designed for sparse user interaction data on digital tourism platforms (DTPs), where conventional tourism recommendation systems often fail to perform effectively. Our approach is based on advanced user profiling through extensive feature transformations, deep neural profile learners, intelligent embeddings, and clusters with associative similarity. The framework presents a ranked list of TE recommendations to users with predicted ratings and sentiments. For enhanced usability, our framework augments recommendations with contextual features: weather forecasts, live traffic updates, and multi-modal interaction tools —including chatbot and voice recommendations. Comparative evaluations with baseline models reveal that our model shows decreased MSE by percentages 38.00%, 61.03%, and 19.48% for Airbnb, TripAdvisor, and Booking.com datasets, respectively. It also outperforms baseline models for MAE, MAAPE, and R $$^2$$ parameters. In addition to utility error measures, our framework demonstrates increased diversity by 0.53%, 0.55%, and 0.35%, and exhibits better item space coverage by 30.12%, 55.02%, and 42.93% for Airbnb, TripAdvisor, and Booking.com datasets, respectively. This study advances digital tourism with practical implications, specifically for sparse-interaction data scenarios.