Simon Gottschalk
Research Group Leader at L3S Research Center, Leibniz University Hannover
Current Projects
In the soofi (Souvereign Open Source Foundation Models) project, a larger AI language model is being developed and made available as open source for use in business and society. Based on a large language model (LLM), a so-called reasoning model is also being created using specialized methods to improve the overall system’s quality and optimize resource consumption. In addition, initial use cases are to be implemented using AI agent technologies.
The European Digital Innovation Hub for AI and Cybersecurity (DAISEC) supports companies from the production, mobility and skilled crafts sectors as well as the public sector in Lower Saxony in the application and development of Artificial Intelligence and Cybersecurity.
In the project KG4MBSE, knowledge graphs (KGs) are used to work on central problems in model-based systems engineering (MBSE). MBSE is being used more and more frequently to counter the increasing complexity and interdisciplinarity in development projects. In early development phases, extensive system models with a comparatively high degree of formalisation are created to integrate different views of the system to be developed as well as the networking of individual model elements such as requirements, system elements or stakeholders. The aim of KG4MBSE is to reduce modelling effort and increase model quality through the targeted reuse and integration of model fragments using knowledge graphs.
Understanding mobility behavior, for instance, traffic flows or public transportation usage in urban areas, has become an increasingly important challenge for a wide variety of stakeholders, including social and environmental scientists, urban planners, federal statistical agencies, and policymakers. These stakeholders are particularly interested in understanding the short-, mid-, and long-term evolution of mobility behavior in specific geographic regions and its interdependencies with other factors. Such research requires high-quality data regarding the actual mobility behavior, as for example, historical archives of vehicle traffic data, shared bike usage, and public transportation usage.
RUSHMORE (2025-2028) will address sparsity and costliness of mobility data through representative synthetic data generation and specialised machine learning models for spatio-temporal knowledge transfer. The project aims to develop a publicly available data service that offers mobility data and metadata to interested data consumers according to the FAIR principles.
Individual travel planning and the regional control of travel flows require innovative algorithms and data diversity. In order to enable attractive, personalised, environmentally aware and cross-modal travel offers, a major opportunity for mobility and tourism providers lies in capturing user needs in a dialogue-based manner and creating AI-based offers to reduce the time required for travel planning and to optimally utilise the infrastructure. The MoToRes project (2024-2026) will explore the potential of a variety of heterogeneous data and develop new AI algorithms, the MoToRes platform for technical orchestration and prototypes for end users.
- My role: Project leader.
Former Projects
The ATTENTION! project (2022-2025) project will systematically collect, model and analyse multiple global data sources such as trade data for imports and exports, company and web data. Based on global trade activities and their contextual information, artificial intelligence (AI) models will be developed and applied to detect and uncover illicit trade activities and their patterns in global, heterogeneous data. Among the challenges to be overcome are the lack of known cases, the complexity of the patterns and the need to explain detected suspicious cases. To this end, ATTENTION! relies on the interplay of supervised and unsupervised learning as well as the creation of a knowledge graph.
- My role: Project leader of the German consortium.
Die zunehmende Digitalisierung in vielen Anwendungsbereichen der Wirtschaft, Verwaltung und Wissenschaft führt zu ständig wachsenden Datenmengen und damit immer häufiger zur Notwendigkeit, datenbasierte Prognosen zu erstellen und Zusammenhänge in großen heterogenen Datenmengen zu erkennen. Machine Learning (ML) ist ein Kernthema in diesem Bereich, mit dem sich 64% aller deutschen Unternehmen aktiv beschäftigen. Die effiziente Anwendung aktueller ML-Verfahren erfordert jedoch ein sehr hohes Maß an Expertenwissen, was einer verbreiteten Nutzung von Machine Learning-Ansätzen, insbesondere durch kleine und mittlere Unternehmen (KMU), im Wege steht.Die zentrale Forschungsfrage des Simple-ML-Projekts (2018-2022) lautet daher: Wie kann die Benutzbarkeit von ML-Verfahren signifikant verbessert werden um diese für einen breiteren Anwenderkreis leichter zugänglich zu machen?
- My role: WP leader and project leader.
The ALEXANDRIA project (2014-2019) (ERC Nr. 339233) aims to develop models, tools and techniques necessary to explore and analyze Web archives in a meaningful way. ALEXANDRIA will significantly advance semantic and time-based indexing for Web archives using human-compiled knowledge available on the Web, to efficiently index, retrieve and explore information about entities and events from the past. The ALEXANDRIA Testbed will provide relevant collections and algorithms that enable further research on and practical application of research results to existing archives.
- My role: Project member (2015-2019).