Making social history research papers machine interpretable

Supervisor: Lise Stork (


Transparency and clarity of scientific hypotheses on which we build is paramount. Natural language cannot carry truth unambiguously, but scientific discoveries are commonly described, albeit with academic language, in natural language in scientific articles. The formalisation of hypotheses advances research: it facilitates transparency (e.g., for reproducibility), it can help highlight bias, as well as theory synthesis and thereby the revision, consolidation and the unification of known theories.


The goal of this project is to see if we can partially automate or support the construction of knowledge graph on social history hypotheses from the literature.


We intend to use an NLP tool called “construction grammar” to extra semantic frames from the text of the meta-review. The first challenge is to semi-automatically recognise elements in the scientific papers that are relevant for the hypothesis. This should make it easier for experts to construct the hypothesis graph based on these suggestions. The second challenge would be to generate complete hypothesis graphs using the construction grammars, and to evaluate the output with human experts.


supervision will be by Lise Stork. We will collaborate with a team in Brussels who built the construction grammar, and with social historians from the International Institute of Social History (IISG).