A Modular Social Sensing System for Personalized Orienteering in the COVID-19 Era

Abstract

Orienteering or itinerary planning algorithms in tourism are used to optimize travel routes by considering user preference and other constraints, such as time budget or traffic conditions. For these algorithms, it is essential to explore the user preference to predict potential points of interest (POIs) or tourist routes. However, nowadays, user preference has been significantly affected by COVID-19, since health concern plays a key tradeoff role. For example, people may try to avoid crowdedness, even if there is a strong desire for social interaction. Thus, the orienteering or itinerary planning algorithms should optimize routes beyond user preference. Therefore, this article proposes a social sensing system that considers the tradeoff between user preference and various factors, such as crowdedness, personality, knowledge of COVID-19, POI features, and desire for socialization. The experiments are conducted on profiling user interests with a properly trained fastText neural network and a set of specialized Naïve Bayesian Classifiers based on the “Yelp!” dataset. Also, we demonstrate how to approach and integrate COVID-related factors via conversational agents. Furthermore, the proposed system is in a modular design and evaluated in a user study; thus, it can be efficiently adapted to different algorithms for COVID-19-aware itinerary planning.