Graph-Oriented Models for Integrating, Mapping, and Querying Open Data for Sustainability
By Wissal Benjira, my former PhD student co-advised with Benedicte Bucher, Faten Atigui et Malika Grim-Yefsah – Co-funded between IGN & DVRC-ESILV
[theses.fr, manuscrit]
One of the most pleasant PhD thesis as a supervisor. Always listening, hard working, and smart.
She succeeded in creating a bridge between geomaticians and computer scientists (especially on graph databases ;). Great results with VLDB, DKE, IEEE Big Data publications and Best demo Award at BDA’25!
The Sustainable Development Goals (SDGs) defined by the United Nations constitute a global ref- erence framework that specifies targets related to social, environmental, and economic sustainability, together with indicators intended to measure progress toward these targets. These indicators corre- spond to quantitative measures whose values must be computed at multiple territorial scales, ranging from the national level to finer levels such as regions or municipalities.
While some indicators can be computed using data produced by dedicated institutional bodies, such as INSEE or IGN in France, others do not rely on directly exploitable data sources. In this context, open data represents an important opportunity. However, despite the abundance of such data, its use to compute reliable, comparable, and reproducible indicators for specific territories remains a major challenge. Data heterogeneity, the lack of semantic harmonization, and the limited transparency of computation methods reduce trust in and interpretability of the resulting values.
Therefore, this thesis relies on graph-oriented models for the integration, mapping, and querying of open data to support the computation of SDG indicators. The proposed approach is based on a dual-graph architecture, combined with the use of language models. A domain graph, denoted SDG Graph, models the entities, attributes, and relationships associated with indicators as described in the official documentation of the United Nations Agenda. In parallel, a denoted Metadata Graph, describes open data sources. The alignment of these two graphs forms a knowledge graph, denoted SDG-KG, which allows joint reasoning over indicators and the data required for their computation.