The conference

The first African Data Science Conference (ADSC 2026) was held at the Wits Science Stadium, University of the Witwatersrand, Johannesburg, from 24 to 26 June 2026 under the theme African Data Science, for Africa. Wits University describes it as a continental forum connecting scientific research, practice, policy and capacity building in data science and artificial intelligence.

Metraltus contributed the oral presentation Computational Food Identity: A Multidimensional Data Model for Algorithmic Food Certification. The published abstract is the public point of origin of the Metraltus research programme on multidimensional food representation, algorithmic certification and food knowledge graphs.

Published abstract

Reproduced verbatim from the official ADSC 2026 programme.

Food certification systems such as geographical indications and quality labels rely on fragmented administrative processes and monodimensional representations of food products, limiting their integration into modern data-driven infrastructures. This paper introduces a data science framework for representing food products as multidimensional computational entities. We propose the concept of Computational Food Identity, implemented through a multidimensional representation integrating geographic, biological, cultural, environmental, social, and legal attributes. The proposed representation, referred to as the Multidimensional Food Digital Twin (MFDT), transforms heterogeneous food data into structured feature vectors encoded as tensor representations. Based on these representations, we introduce a computational certification mechanism in which dimension-specific scoring functions generate algorithmic micro-certifications associated with different attributes of the product. The framework also supports the construction of a food knowledge graph linking products, producers, territories, and ecosystems, enabling relational analysis and potential graph-based machine learning applications. A prototype implementation illustrates how heterogeneous food data can be encoded within the MFDT representation and used to generate multidimensional certifications informed by a mixed-method field study conducted in Grand-Bassam, Côte d’Ivoire, among small-scale women attiéké producers to assess the feasibility and potential impact of a smartphone-based application for direct multidimensional data collection.

Scientific significance and limits

The proposal treats a food product as a situated entity rather than a single label or database record. It connects formal representation, evidence, territory, producers and ecosystems. This creates a research path toward traceable multidimensional claims and relational analysis while making the assumptions of certification systems more explicit.

The abstract presents a framework and prototype. It does not by itself establish regulatory validity, generalisability across food systems or a deployed certification service. Those questions require further empirical, legal, technical and participatory validation.

Suggested citation

Galliano, J. F. (2026). “Computational Food Identity: A Multidimensional Data Model for Algorithmic Food Certification.” Oral presentation and published extended abstract, African Data Science Conference 2026, University of the Witwatersrand, Johannesburg, 24 June 2026. No DOI assigned.