Geographic entities Evolution in Knowledge graphs – GEvoK

The project is structured into five complementary work packages (WPs).

WP0 focuses on project management and the dissemination of results (organization of kick-off and final seminars, overall coordination, and team management).

WP1 aims to develop a generic ontological model to describe territorial changes and, in a second phase, to build a software tool that enables its implementation using domain-specific vocabularies defined by expert communities. For example, the generic concept of “expansion” may be specified as “urbanization” in urban studies.

WP2 is dedicated to the generation of spatio-temporal knowledge graphs (ST-KGs) describing territorial trajectories. Its main objective is to extend existing change detection and description algorithms to new types of spatial objects (points, lines, regular grids), and to integrate new methods based on deep and machine learning.

WP3 focuses on enriching and querying the knowledge graphs produced in WP2. This includes the use of large language models (LLMs) for natural language querying and for contextualizing changes (through a semi-automatic approach validated by experts). It also involves linking study objects to encyclopedic knowledge graphs such as Wikidata and DBpedia, and computing metrics to quantify changes and support analytical scenarios (e.g., comparing the evolution of different vegetation areas over time under varying pressures).

WP4 is dedicated to experimentation and validation. Two case studies related to climate change will be conducted: the transformation of cities—particularly Grenoble—to assess the impact of low-carbon mobility policies (data sources: OpenStreetMap and metropolitan open data portals), and the evolution of mountain pastures, which are highly affected by climate change and human activity, to support public authorities in their preservation efforts (data sources: Landsat and Sentinel satellite imagery, and CORINE Land Cover).

The tool, based on linked open data, will be made available to citizens during an experimental phase.

The major expected outcomes are:

1. The creation and publication on the Linked Open Data Cloud of a generic ontology describing territorial changes, along with associated application ontologies applied to specific domains.

2. A tool-supported methodology for detecting and describing territorial changes, leveraging the developed ontological models as well as conflation, change detection, and trend analysis algorithms from machine or deep learning.

3. ST-KGs (enriched with information from other KGs or LLMs) describing territorial changes in the city of Grenoble and alpine pastures over the past decade. Published as Linked Open Data on the Web and based on the generic ontology, these ST-KGs will be queryable in natural language.

The GEvoK project represents a major scientific advance by proposing an innovative, automated, and standardized software solution to manage the evolution of geographic data over time. In the field of knowledge representation, it addresses a significant gap, as no existing model or tool currently enables the detection and representation of complex territorial changes with such a level of generality.

This approach opens new perspectives for a wide range of scientific communities by providing tools capable of automating the analysis and description of territorial dynamics. For example, in ecological engineering, GEvoK will facilitate the study of invasive species spread, while in urban planning, it will enable comparisons of structural and morphological changes in cities.

By making its model and tools available as open data and open source, the project will allow researchers from various disciplines—such as history, environment, agriculture, and ecology—to reuse and adapt this framework for their own work, thereby fostering stronger interdisciplinary collaboration.