As each year, I am looking for talented and driven master students to work on a variety of projects out of the ICAI OpenGov Lab. See the one-sentence summaries (provided by ChatGPT) of the project proposals below! And reach out if you’re interested in working on any of these (and want to see the full project description);
- Labeling and Detecting Mis- and Disinformation in OpenGov Data
Develops methods to detect known false claims in government texts using weak supervision and fact-check datasets, to study how mis- and disinformation surface, spread, or are contested in governmental discourse. - Quantifying Source Authoritativeness from Governmental Data
Explores how citation and reference patterns in parliamentary documents can be used to measure the authority of sources, using graph-based metrics and debate uptake as signals. - Homophore Resolution in Governmental Texts
Investigates automatic detection and resolution of generic government-specific references (“het ministerie,” “de commissie”) by combining general and domain-specific knowledge bases. - NLI for Measuring Political Party Program-Parliament Consistency
Uses Natural Language Inference between party programs and parliamentary debates to quantify how consistently parties and politicians adhere to their stated programs over time. - VLMs and Document AI for Open Government Data
Applies multimodal models to segment and classify heterogeneous PDFs (e.g., merged reports, emails, chat logs) into meaningful components, improving search, summarization, and retrieval. - Large-Scale Topic Modeling and Narrative Detection for Parliamentary Proceedings
Applies modern topic modeling and clustering techniques to detect themes, narratives, and framing strategies in parliamentary debates, with potential visualization of evolving discourse. - Bias and Accessibility in LLM-Based Summaries of Open Government Documents
Studies bias in AI-generated summaries of government texts, examining omission or misrepresentation risks and exploring fairness metrics and accountability tools. - Computational Linguistic Analysis of Government Documents
Analyzes the linguistic characteristics of Dutch governmental legal language computationally, to inform simplification models and quantify differences from plain language. - Decision Models and Open Government Documents
Explores links between decision models (e.g., DMN) and textual decision documents, aiming to automate text-to-model and model-to-text transformations for improved explainability and retrieval. - Recognising Official Entities in Government Documents
Develops specialized NER models to detect government-specific entities (committees, legal references, institutions) beyond standard KB coverage, enhancing linking and retrieval. - To delete or to archive? Classification of Government Documents against Selection Lists
Examines automatic classification of government documents against archival selection lists to support decisions on deletion versus long-term storage, reducing manual effort.