Feeding a Gazetteer

We propose a workflow for retrieving place names from a corpus of Hebrew historical newspapers. We show that using an initial curated set of unambiguous toponyms and vector similarity is a productive method to populate a gazetteer with previously unknown variant names of known places, as well as names of places which are not yet included in the Gazetteer. We examine several parameters for enhancing accuracy and suggest a workflow that combines computation with human expertise, and is valuable to spatial history as well as to other domains.