Advanced Methods for Semantic Enrichment and Summarization of Mobility Data

Mobility data from sources like GPS devices, mobile networks, and Internet of Things sensors offers valuable insights into the movement of people and objects, but it often lacks meaningful context. The first thesis contribution is a novel methodology for the semantic enrichment of raw spatio-temporal data, i.e., trajectories, allowing for the integration of various contextual elements, such as points of interest (POIs), transportation modes, and environmental factors, to name a few. The semantic enrichment provides deeper insights into movement patterns and behaviors. However, the complexity and heterogeneity of semantic trajectories present challenges for storage and analysis. To face these challenges, the second thesis contribution is to propose a summarization method that reduces the volume of enriched trajectories while preserving essential semantic information. This approach involves segmenting geographical areas into meaningful regions and discretizing trajectories based on the regions traversed. Dividing areas into semantically relevant regions raises several challenges regarding their identification and representation, also considering new and richer data sources. The third thesis contribution is a dynamic method for identifying and representing urban regions using user-generated data, such as social media posts and geolocated images. The approach leverages Large Language Models (LLMs) to produce human-readable summaries, enhancing the interpretation of the regions and giving a human-centric perspective of the urban environment. The contributions of this thesis lie in offering robust tools for enriching, summarizing, and analyzing mobility data, addressing key challenges related to data volume, heterogeneity, and context-awareness.