In order to enhance tourists’ experiences, Destination Management Organizations need to know who their tourists are, their travel preferences, and their flows around the destination. The study develops a methodology that, through the application of Artificial Intelligence techniques to social media data, creates clusters of tourists according to their mobility and visiting preferences at the destination. The applied method improves the knowledge about the different mobility patterns of tourists (the most visited points and the main flows between them within a destination) depending on who they are and what their preferences are. Clustering tourists by their travel mobility permits uncovering much more information about them and their preferences than previous studies. This knowledge will allow DMOs and tourism service providers to offer personalized services and information, to attract specific types of tourists to certain points of interest, to create new routes, or to enhance public transport services.

Tourists who visit a city for the first time may find it difficult to decide on places to visit, as the amount of information in the Web about cultural and leisure activities may be large. Recommender systems address this problem by suggesting the points of interest that fit better with the user’s preferences. This paper presents a novel recommender system that leverages tweets to build user profiles, taking into account not only their personal preferences but also their travel habits. Association rules, which are mined from the previous visits of users documented on Twitter, are used to make the final recommendations of places to visit. The system has been applied to data of the city of Barcelona, and the results show that the use of the social media-based clustering procedure increases its performance according to several relevant metrics.

It is a known fact that the order in which touristic activities are experienced plays a role in how enjoyable they are. This is the reason why tourists prefer to book carefully prepared day tours on arrival to a new destination, as they allow them to see the essence of the destination while traversing scenic routes. Tours are great, but they are expensive, do not allow room for personal exploration, and are built as a one-size-fits-all which does not consider the individual preferences of the tourist. In contrast, it is possible to make an optimal selection and ordering of touristic activities from a larger set of possibilities that match a tourist’s personal preferences, balancing important aspects like diversity, spatial proximity, or degree of interest on popular places. We propose a multi-objective genetic algorithm that uses a weighted averaging operator to balance four diverse objective functions crafted to maintain diversity, proximity, interest on popularity, and cultural preference. The system has been evaluated against four baseline algorithms and found to perform significantly better for the specified purpose.

Analysing social media messages shared during a climate disaster to extract relevant information is a difficult task due to the high volume of messages and the subjective nature of relevance. Nevertheless, this is a necessary step when annotating texts to train models for crisis management. Traditionally, samples are annotated manually, which is time consuming and challenging to maintain consistency, especially when dealing with multilingual corpora. Instruction-tuned generative language models are a potential solution to this problem due to their ability to follow instructions and their state of the art performance in several natural language processing tasks including text classification. This paper evaluates several open-source multilingual instructed models with the goal of seeing how well they perform as annotators or text filtering tools in a crisis. We conduct experiments using both zero-shot and few-shot prompting, exploring different ways of incorporating examples in few-shot prompting. Finally, we examine the impact of post-annotation label cleaning and further refine results through score thresholding.