The food system is a complex network encompassing various stakeholders, including primary producers, manufacturers, retailers, and consumers. Across all the stages of the food supply chain, a significant amount of data is produced, ofering valuable insights crucial for ensuring the delivery of safe, high-quality, and cost-efective products to meet the needs of a growing global population. Recommender systems are commonly used in the food domain but often lack personalization, leading to generic recommendations. Enhancing user experience through explainability ofers transparent reasoning behind recommendations, fostering trust and informed decision-making. Semantic reasoning can be enhanced through ontology-based user profiles. Moreover, the increased data sharing in the food sector has raised privacy and security concerns, prompting the development of privacy-preserving data platforms. This PhD project aims to address these challenges by (1) utilizing ontologies for enhancing semantic interoperability, (2) employing eXplainable Artificial Intelligence (XAI) methods and semantic reasoning for enhancing the transparency of recommender systems, and (3) designing a privacy-preserving data platform that facilitates data sharing while ensuring the protection of sensitive information.
The food system comprises a multitude of participants across the supply chain: (1) primary producers
such as cultivators, farmers, and aquafarmers, (2) food producers such as breeders, processors, and
packers, (3) retailers such as distributors, supermarkets and restaurants, (4) logistics such as transporters,
shippers and carriers, and (5) consumers [
A guidance for achieving appropriate data management is the FAIR principles [
In the food domain, recommender systems are commonly used for recommending diferent food
items or recipes in accordance with the user’s preferences, nutritional goals, or dietary restrictions.
Current food recommendation systems face several limitations such as the lack of personalization
by often relying on generic suggestions rather than taking into account user’s preferences, dietary
constraints, cultural heritages, and health-related needs [
Enhancing user experience can be achieved through explainability, which ofers transparent and
understandable reasoning behind specific recommendations [
XAI can be implemented by using post-hoc approaches which refer to the application of XAI models
on top of the black-box output generated from a ML model [
The increase of data sharing among the actors and sharing of personal data and health data has led to
increased awareness of privacy and security concerns. Privacy-preserving data platforms are systems
that facilitate data sharing while ensuring the protection and security of sensitive information [11, 12].
Such platforms enable users to make well-informed food decisions while safeguarding their sensitive
data, thereby enhancing the security and reliability of the user experience [
This PhD project aims to dive into these areas: (1) the usage or design of ontologies for enhancing semantic interoperability in the food supply chain, (2) the usage of XAI methods and semantic reasoning for explaining the decision-making of the recommender systems and the generated output, and (3) the design of a privacy-preserving data platform.
The development of this future food system inspires the following Research Questions (RQ) that will be carried out throughout the PhD.
The main RQ is:
How to design a next-gen cross-sectoral data platform that ensures semantic interoperability, privacy-preserving, and provides Explainable AI solutions?
The sub-research questions are: 1. How to enhance semantic interoperability and explainable artificial intelligence (XAI) in food supply chains or food recommender systems? 2. What privacy-preserving techniques are most efective for protecting sensitive agricultural data? 3. What architectural models are best suited for integrating cross-sectoral data while maintaining privacy and interoperability? 4. What are the specific needs and requirements of the various stakeholders in the agricultural sector? 5. What are the potential applications and benefits of the platform in real-world agricultural scenarios?
The methodology used in this PhD project is the Design Science methodology [13]. The Problem Investigation phase is used to identify what is the state-of-the-art regarding existing ontologies and knowledge graphs in the agri-food domain, XAI techniques that are used in food recommenders, and privacy-preserving techniques. The Treatment Design phase is used to identify (1) requirement analysis, (2) adopting existing web technologies frameworks, (3) designing a semantic web technologiesbased recommender, and (4) modeling of a reference architecture for next-gen privacy-preserving data platform. The Treatment Validation phase is used for (1) validating the used semantic web technologies, (2) using the single case mechanism method in a case study for validating the prototype, and (3) using the expert opinion method for validating the reference architecture models.
Figure 1 illustrates the primary components of the proposed solution, which addresses our research questions in the following manner. The privacy-preserving data platform ensures that each stakeholder retains ownership of their data and can selectively permit access to it. Each data station is managed by a data owner and is accessible via specific algorithms. The platform leverages ontologies to guarantee semantic interoperability. Additionally, the recommender system generates explainable recommendations, thereby assisting users in obtaining informed and transparent outcomes. in AI: Food Recommender System Use Case, In: Proceedings of the 11th International Conference on Human-Agent Interaction, 2023, p. 395-397. [9] M. Rostami, U. Muhammad, S. Forouzandeh, K. Berahmand, V. Farrahi, M. Oussalah, An efective explainable food recommendation using deep image clustering and community detection, Intelligent Systems with Applications, 2022 Nov 1;16:200157. [10] A. Seeliger, M. Pfaf, H. Krcmar, Semantic web technologies for explainable machine learning models: A literature review, PROFILES/SEMEX@ISWC, 2019 Oct 27;2465:1-6. [11] B. Qadeer, M.A. Shah, A. Ishaq, Privacy Preservation in Digital Economy Platforms. [12] A. Durrant, M. Markovic, D. Matthews, D. May, G. Leontidis, J. Enright, How might technology rise to the challenge of data sharing in agri-food?, Global Food Security, 2021 Mar 1;28:100493. [13] R.J. Wieringa, Design science methodology for information systems and software engineering, Springer, 2014 Nov 19.