Construction of a knowledge graph for food health
claims
Remzi Celebi1 , Ilse van Lier2 , Alie de Boer2 and Michel Dumontier1
1
    Institute of Data Science, Faculty of Science and Engineering, Maastricht University, the Netherlands
2
    Food Claims Centre Venlo, Campus Venlo, Faculty of Science and Engineering, Maastricht University, the Netherlands


                                         Abstract
                                         Authorised health claims and their scientific opinions offer valuable insights into health effects of foods
                                         and food ingredients. However, these texts are highly technical and not easily usable to those interested
                                         in the development and use of healthy food products and diets. In this paper, we present our effort to
                                         develop a knowledge graph that was curated from the information of 260 authorised health claims. The
                                         knowledge graph, based on data from scientific opinions, is subdivided into four ontological dimensions:
                                         the food (ingredient); the health effect; the target group; and the scientific evidence underlying the
                                         cause-and-effect relationship. Various differences were found between authorised claims and their
                                         underlying scientific opinions. These findings underline the need for further structuring the approach
                                         to substantiating and assessing health claims. Most importantly however, the development of this
                                         knowledge graph allows consumers, food producers and health care professionals to make personalised
                                         decisions in selecting healthy nutrition.

                                         Keywords
                                         personalised nutrition, food health claim, knowledge graphs, FAIR data




1. Method
We have developed an ontology that reflect the wealth of information found in EFSA’s scientific
opinions on health claims1 , including the specific nutrients or bioactive ingredients, the health
relationships and biomarkers, the conditions of use for such a claim (e.g. the population that is
referred to in the claim) and the supportive evidence underlying these claims.
   To build this ontology, we have reviewed all EU authorised health claims and their underlying
scientific opinions. These opinions detail the active substances (foods or food ingredients),
the beneficial effects as well as the relevant evidence substantiating the relationship between
ingredient and its effect. We have extracted these information manually from all scientific
opinions regarding the authorized claims in the Register. This resulted in the inclusion of 260
claims and their scientific opinions: the scientific substantiation and conditions of use of 235
authorised function claims (229 based on generally accepted scientific evidence and six based
SWAT4HCLS’24: Semantic Web Applications and Tools for Health Care and Life Science Conference, February 26-29,
2024, Leiden, The Netherlands
Envelope-Open remzi.celebi@maastrichtuniversity.nl (R. Celebi); i.vanlier@maastrichtuniversity.nl (I. v. Lier);
a.deboer@maastrichtuniversity.nl (A. d. Boer); michel.dumontier@maastrichtuniversity.nl (M. Dumontier)
Orcid 0000-0001-7769-4272 (R. Celebi); 0000-0001-8381-1252 (I. v. Lier); 0000-0002-6500-4649 (A. d. Boer);
0000-0003-4727-9435 (M. Dumontier)
                                       © 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
    CEUR
    Workshop
              CEUR Workshop Proceedings (CEUR-WS.org)
    Proceedings
                  http://ceur-ws.org
                  ISSN 1613-0073




1
    https://ec.europa.eu/food/food-feed-portal/screen/health-claims/eu-register
Figure 1: The proposed ontology based the Semanticscience Integrated Ontology (SIO) for food health
claims.

The proposed data model for food health claim has four dimensions: Scientific Evidence, Food, Health
  Effect and Target Population. Each dimension is represented by our proposed ontology based on
               Semanticscience Integrated Ontology (SIO) and an existing vocabulary.


on newly developed evidence), 13 disease risk reduction claims and 12 claims on children’s
development and health.
   Our ontology has four sub-structures in order to better explain different dimensions of health
claims, as shown in Figure 1. These relate to the steps taken in assessing scientific dossiers
for health claim authorisations the food or active ingredient itself (here labelled as Food), the
beneficial physiological effect (Health Effect), the potential target group for this beneficial
physiological effect (Target Population) and finally, the scientific evidence that is supports the
association between consuming the ingredient and the suggested beneficial physiological effect
(Scientific Evidence).


Acknowledgments
Research reported in this publication was partially supported by Limburg University Fund /
SWOL. The work of AdB is supported by the Dutch Province of Limburg.