Placenames analysis in historical texts: tools, risks and side effects

International audienceThis article presents an approach combining linguistic analysis, geographic information retrieval and visualization in order to go from toponym extraction in historical texts to projection on customizable maps. The toolkit is released under an open source license, it features bootstrapping options, geocod-ing and disambiguation algorithms, as well as cartographic processing. The software setting is designed to be adaptable to various historical contexts, it can be extended by further automatically processed or user-curated gazetteers, used directly on texts or plugged-in on a larger processing pipeline. I provide an example of the issues raised by generic extraction and show the benefits of integrated knowledge-based approach, data cleaning and filtering