..
We perform foundational and applied research and drive development in Data Science. A particular focus is interdisciplinary work, bridging the gap between generic approaches and specific applications. Application areas in which we are applying our techniques include: Agriculture, Cultural Heritage, Education, Health, Industry 4.0, Legal text analysis, Logistics, Music retrieval, and Social Media analysis.
A further focus is on providing the means to allow us to trust the complex socio-technical systems we create and interact with. Our motto, Insights for Society, highlights our consideration of the wider consequences for society of our research, such as climate change related to computationally intensive tasks, filter bubbles caused by recommender systems, and the effects of decision support on everyday work.
Our research is summarised in the diagram. We describe each of the four main areas in more detail:
Analysis has the goal of discovering or extracting useful information from data for informing conclusions. The data can be in the form of natural language (Natural Language Processing), music or numerical data (Data Mining). Semantic Systems bring pre-existing knowledge in the form of ontologies and related structures into the analysis. The data can also be mathematically modelled, for example through machine learning. Simulations can be run to investigate the outcomes based on different parameterisations of the models.
Interaction deals with how people are supported in their interactions with large repositories. For Information Retrieval, the user provides a query used to retrieve relevant objects from the repository. Recommender Systems retrieve relevant objects based on previous user interactions with the system. With Conversational Systems, such as chatbots, getting to the required information should be through a natural language conversation with a system. Many tasks in both the Analysis and Interaction areas need to learn from large amounts of manually annotated data - here we work on approaches to make the generally time-consuming process of manual annotation more efficient.
Infrastructure research deals with supporting various aspects of doing Data Science. Secure Data Infrastructures allow calculations to be done on data without the data itself being shared. Research Data Management ensures that data remains findable, accessible, interoperable, and reusable indefinitely. Virtual Research Environments facilitate online collaboration of researchers.
We ensure that our research has Impact on society. Our approaches to support Reproducible Research contribute to overcoming the reproducibility crisis in multiple areas of science. Our analyses form the basis for Evidence-Based decisions in industry and government. Our work on transparency of AI models and their outputs contributes to developing Trustworthy AI. The Digital Humanism initiative ensures that technology development remains centred on human interests.
Research domains within the Research Unit include: