Postdoctoral Researcher in Transformative Geo-analytical Question Answering

Postdoctoral Researcher in Transformative Geo-analytical Question Answering

Utrecht University - Faculty of Geosciences

Utrecht, Netherlands

We are looking for a highly motivated postdoctoral researcher in Geoinformatics, Artificial Intelligence, GeoAI, Ontology Engineering to join the ERC-funded project GeoTrAnsQData at Utrecht University. The project aims to develop theory as well as novel methodology for transformative geographic question answering based on conceptual transformation graphs, semantic workflows, and AI-based reasoning over geodata repositories.

Your job

Answering spatial questions with geodata is a core task across domains like urban planning, sustainability, and public health. Yet, current Geographic Question Answering (GeoQA) tools typically return factoid answers, datasets or maps rather than help users construct meaningful spatial analyses.

Most geographic questions data scientists may want to answer require some form of geo-analytical transformation of maps. Answer maps need to be (re)generated from data, rather than retrieved from storage. The GeoTrAnsQData project addresses this challenge by developing a novel GeoQA methodology that interprets geographic questions into executable geo-analytical workflows which transform geodata repositories into possible answer maps. In the project, we develop knowledge graph models of such transformations that are self-describing when applied to existing map repositories, and can be scaled up to large data repositories using state-of-the-art AI methods. In this way, GeoQA becomes independent from the data stored in a particular database and instead allows users to explore the space of possibilities of reusing maps for various purposes as specified in a question.

You will be leading the development of the core computational mechanism for this novel GeoQA system. In this project, you will:

• design and test conceptual or cognitive models for geodata sources and geo-analytical transformations;
• design and implement algorithms that parse geographic questions into conceptual transformation graphs;
• develop graph-based methods (e.g. knowledge graph embedding, graph neural networks) to match questions to relevant geodata workflows;
• integrate semantic web technologies (e.g. RDF, SPARQL, OWL), large language models and graph neural networks into the reasoning pipeline;
• build a prototype system that supports interactive workflow synthesis and question answering;
• collaborate with a technical assistant and two PhD candidates working on geodata/workflow modelling and question parsing;
• co-supervise junior researchers and contribute to scientific publications and open-source tooling.

Your qualities

• A PhD in Computer Science, Artificial Intelligence, GeoInformatics/GIScience, Geocomputation or a related field;
• strong programming skills and experience with knowledge graphs, semantic web standards, or graph-based reasoning;
• a strong interest in conceptual modeling, and experience in semantic modeling, ontology engineering, formal logics or cognitive modelling;
• affinity with maps, geospatial concepts, GIS systems, and spatial data infrastructures;
• experience with AI and machine learning methods, especially in the areas of natural language processing or graph neural networks;
• the ability to work independently and collaboratively in an interdisciplinary environment;
• excellent English writing and communication skills;
• experience with leading projects, mentoring or supervising students is a plus.

Apply NowDeadline 16 July