Unlock how geographic data can answer complex societal questions: in this PhD you develop semantic and AI-driven models that transform diverse geodata into actionable answer maps. Join our ERC-funded GeoTrAnsQData project and help shape the future of geographic question answering.
Your jobGeographic questions like 'What is the potential to reduce urban heat in Amsterdam by installing green roofs on existing buildings?' are important in fields such as urban planning, sustainability, and public health. It requires the transformation of maps combining different suitable geodata sources, including heat sources and building layouts to generate an answer map. This problem is called indirect question answering, and it is not straightforward with current GeoQA tools.
In such scenarios, maps must be created or transformed from other maps. The ERC funded project GeoTrAnsQData project addresses this by developing a GeoQA method that converts questions into executable geo-analytical workflows, turning geodata into new answer maps. Knowledge graphs can be used to model these transformations and to link geodata sources to questions. In this project we will apply symbolic and sub-symbolic AI methods to scale this up across large geodata repositories, enabling users to explore various ways maps can be reused to answer different kinds of questions. We are looking for a motivated PhD candidate to conduct interdisciplinary research that links question answering, knowledge modelling, geo-spatial analysis, and workflow construction. This PhD position focuses on developing a semantic model of the diverse geodata sources in a map repository, enabling reasoning about their analytical purposes and their provenance using concepts of geographic information. You will focus on modelling data resources and workflow generation for geographic question answering.
You will:
- develop a semantic framework and knowledge graph to represent geodata sources, their analytical purpose and their provenance, including abstract geospatial workflows;
- design AI- and machine-learning-based methods that automatically describe and model geodata sources using textual metadata (NLP) and the geodata itself;
- contribute to a corpus of geo-analytical scenarios with questions and corresponding workflows;
- collaborate closely with another PhD candidate (question modelling), a postdoc (GeoQA reasoning engine) and a technical assistant;
- evaluate your framework through user-centred scenarios in spatial-planning or environmental-assessment contexts, and develop annotation manuals and gold standards for benchmarking.
This role is ideal if you are excited about conceptual modelling, spatial data infrastructures and AI-supported reasoning over geographic information. You will work at the Department of Human Geography and Spatial Planning, contributing to cutting-edge research on spatial reasoning, semantic technologies and AI for geosciences.