Extracting Causal Claims from Information Systems Papers with Natural Language Processing for Theory Ontology Learning

摘要

The number of scientific papers published each year is growing exponentially. How can computational tools support scientists to better understand and process this data? This paper presents a software-prototype that automatically extracts causes, effects, signs, moderators, mediators, conditions, and interaction signs from propositions and hypotheses of full-text scientific papers. This prototype uses natural language processing methods and a set of linguistic rules for causal information extraction. The prototype is evaluated on a manually annotated corpus of 270 Information Systems papers containing 723 hypotheses and propositions from the AIS basket of eight. F1-results for the detection and extraction of different causal variables range between 0.71 and 0.90. The presented automatic causal theory extraction allows for the analysis of scientific papers based on a theory ontology and therefore contributes to the creation and comparison of inter-nomological networks.