Student research project
Can AI extract light interventions from papers?
Evaluating NLP and information extraction for research synthesis.
The question
Research reports describe light interventions in inconsistent ways. Could AI help researchers locate and organise relevant information while keeping uncertainty and source evidence visible?
The project
Explore natural language processing and information extraction from scientific papers. The aim is to assess where automated assistance can support evidence synthesis and where human judgement remains necessary. Detailed extraction specifications and evaluation design will be developed within the project.
What you will do
- Review information-extraction approaches and reporting challenges in light-intervention research.
- Build a reproducible prototype for an agreed collection of accessible papers.
- Evaluate outputs against human assessment, analysing omissions, unsupported statements and traceability to sources.
Possible scope
A thesis or engineering project could combine prototype development and quantitative evaluation. A shorter placement could analyse a bounded error pattern or workflow component. Strong programming skills and interest in reliable scientific AI are useful.
How to apply
Send a short motivation, CV, relevant skills, preferred project, dates, format and location. Mention external funding. Contact Prof. Manuel Spitschan using the address below.
Location, timing and scope are agreed with the supervisor; availability depends on capacity and the work involved.