How do different computational techniques construct the keyword? And how can we understand keywords as map-maping techniques of cultural phenomena?
In a new paper for Digital Humanities Quarterly, Tobias Blanke and Chloe Papadopoulou investigate the continued importance of keywords in digital humanities and especially their relation to recent machine-learning approaches. They take three approaches – TF-IDF, KeyBERT, and attention – to extract keywords from thousands of UN documents on security and freedom. Among other things, they find that the extracted keywords differ significantly and demonstrate how to tell ‘keyword stories’ with the different techniques.