POIs are complex and evolving entities exhibiting complex and evolving relations. Data about POIs are produced and collected by various stakeholders, with various methods, and for various purposes, using different processes, representation models and formats. Thus, POI data are highly fragmented, coming from diverse and heterogeneous sources, with significantly varying complexity, quality, completeness and timeliness. Moreover, they lack common identifiers, making it very challenging to identify and assemble pieces of information for the same POI from different sources.
SLIPO will address the limitations, gaps and challenges of the current landscape in collecting and integrating POI data. Specifically, our approach focuses on developing tools and services that will tackle the following challenges:
- Lack of standardization in data models, formats and identifiers. Despite the fact that POI data are ubiquitous, there are yet no de jure standards in models, formats and identifiers. As a result, POI datasets provided by different vendors are often not compatible with each other and require excessive effort and domain knowledge to be integrated and reused.
- Inherent ambiguity of POIs. POIs are entities having a twofold nature, geospatial and semantic; moreover, their characteristics and associated information evolves over time. This results in multiple sources of ambiguity when dealing with POI data. For example, the same POI may appear with slight naming variations in different sources, while different POIs may in fact have the same or similar names.
- Long update cycles. Due to the effort needed for maintaining and curating POI datasets, the information contained is typically relatively static, focusing on factual aspects of a POI (e.g. title, category), and ignoring the evolution and provenance of POIs. Hence, data are rarely updated and, often, there are no historical profiles of POIs that evolve over time and keep track of associated events.
- Fragmented and disconnected POI profiles. Based on how and for what purpose a POI dataset has been created, the contained information will typically cover only certain aspects of the POIs. For example, a navigation service and a city guide may have different priorities when deciding which POIs to include and what kind of information about them to collect.
- POIs treated independently and out of context. Last but not least, POI datasets are typically treated as collections of individual entities. Each POI is modeled, stored and analyzed independently, without considering or establishing connections and links to other POIs. This significantly limits the type of analyses that can be carried out over the sets of POIs and creates gaps with the actual needs of users.