Finding structure in multi-modal food data with visual analytics

Descrizione

Finding structure in multi-modal food data with visual analytics Sintesi in italiano (traduzione automatica): Questa posizione di dottorato fa parte del progetto HSFood4ALL, un'iniziativa di ricerca interdisciplinare che si concentra su alimentazione sana e sostenibile. Il candidato lavorerà su analisi esplorativa dei dati complessi, integrando informazioni ambientali, nutrizionali e sociali per sviluppare strumenti di etichettatura innovativi e strategie di comunicazione. Le mansioni principali includono l'analisi dei dati attraverso tecniche di machine learning non supervisionato e la progettazione di rappresentazioni visive efficaci per facilitare decisioni informate da parte di consumatori e policy maker. È richiesta una laurea magistrale in Data Science, Ingegneria Biologica, Informatica o un campo correlato. La sede di lavoro è presso l'ente di ricerca coinvolto nel progetto. This PhD position is part of the HSFood4ALL (Healthy and Sustainable Food for All) project. HSFood4All is an ambitious interdisciplinary research project that brings together expertise in agricultural and environmental economics, nutrition, consumer behaviour, data science, sustainability assessment, governance, and participatory research. Despite growing awareness on sustainability and healthy diets, dietary patterns only improve slowly due to a lack of incentives and facilitators for consumers. Information is fragmented, labels are often confusing, and food environments frequently make healthy and sustainable choices difficult. At the same time, food producers and retailers struggle to communicate sustainability efforts in ways that are meaningful and trusted by consumers. The HSFood4All project addresses these challenges by: Integrating environmental, nutritional, social, and economic data. Developing innovative multi-dimensional food labelling and communication tools. Studying consumer behaviour, food literacy, and food environments. Testing behavioural interventions and communication strategies. Co-creating solutions through citizen research labs and stakeholder engagement. Delivering evidence-based recommendations for industry and policymakers This particular PhD position sits in a transversal work package that provides the analytical and visual foundation for the whole project, working across its three facets: healthy food choices, environmental sustainability, and voluntary sustainability standards. This cross-cutting position is where the research questions come from. Integrating data across facets raises hard problems in its own right, and the patterns worth finding are those that appear only once the facets are joined. The candidate works at that intersection and builds an independent research line from it. The work is exploratory data analysis in the detective sense: searching complex, multi-modal data for clues, and letting the structure that emerges drive the next question. Three connected threads make up the research: Discovery: what patterns hide across the facets? - Using unsupervised machine learning (clustering, dimensionality reduction), topological data analysis and interactive visualisation in a tight loop with domain experts, surface structure and cross-domain relationships (nutrition–environment trade-offs, consumer segments, cross-facet dependencies) that siloed analyses cannot reach. Which methods reveal which structure, and where they agree or diverge, is open. Integration: how do you make heterogeneous data speak to each other? - Sources differ in resolution, structure, and completeness, and the choice of representation (relational, document or graph database) determines which cross-modal structure can be used for downstream unsupervised analysis. Designing that data representation (schema) and the integration strategy over joint consumer/product data is a research problem in its own right. Translation: what makes a visual design effective? - At a later stage in the project, we will design front-of-package (FOP) visual formats at multiple levels of informational detail grounded in perception and cognition, so that consumers and policymakers can make data-informed decisions. What encodings stay faithful to the data while remaining interpretable to non-experts is an empirical, testable question. You will build on AIDA's methodological line (e.g. minimum spanning tree–based topological networks, flare-sensitive clustering, PLSCAN) and extend it, with food systems as the test case. Domain experts across the three facets provide the questions and ground-truth; findings are published at visual-analytics and data-science venues (e.g. EuroVis, IEEE VIS). We are looking for a highly motivated PhD candidate to join our team in this mission to improve food consumption behaviour. Your ideal background: EU Master Degree in Data Science, Bioscience Engineering, Computer Science or a closely related field Demonstr Annuncio in inglese. Fonte: Euraxess (Commissione europea).